Matching Multi-Sensor Remote Sensing Images via an Affinity Tensor
Matching multi-sensor remote sensing images is still a challenging task due to textural changes and non-linear intensity differences. In this paper, a novel matching method is proposed for multi-sensor remote sensing images. To establish feature correspondences, an affinity tensor is used to integra...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2018-07-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | http://www.mdpi.com/2072-4292/10/7/1104 |
| _version_ | 1819279461197021184 |
|---|---|
| author | Shiyu Chen Xiuxiao Yuan Wei Yuan Jiqiang Niu Feng Xu Yong Zhang |
| author_facet | Shiyu Chen Xiuxiao Yuan Wei Yuan Jiqiang Niu Feng Xu Yong Zhang |
| author_sort | Shiyu Chen |
| collection | DOAJ |
| description | Matching multi-sensor remote sensing images is still a challenging task due to textural changes and non-linear intensity differences. In this paper, a novel matching method is proposed for multi-sensor remote sensing images. To establish feature correspondences, an affinity tensor is used to integrate geometric and radiometric information. The matching process consists of three steps. First, features from an accelerated segment test are extracted from both source and target images, and two complete graphs are constructed with their nodes representing these features. Then, the geometric and radiometric similarities of the feature points are represented by the three-order affinity tensor, and the initial feature correspondences are established by tensor power iteration. Finally, a tensor-based mismatch detection process is conducted to purify the initial matched points. The robustness and capability of the proposed method are tested with a variety of remote sensing images such as Ziyuan-3 backward, Ziyuan-3 nadir, Gaofen-1, Gaofen-2, unmanned aerial vehicle platform, and Jilin-1. The experiments show that the average matching recall is greater than 0.5, which outperforms state-of-the-art multi-sensor image-matching algorithms such as SIFT, SURF, NG-SIFT, OR-SIFT and GOM-SIFT. |
| first_indexed | 2024-12-24T00:28:16Z |
| format | Article |
| id | doaj.art-76ede179918e4d88840147835329feaf |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-12-24T00:28:16Z |
| publishDate | 2018-07-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-76ede179918e4d88840147835329feaf2022-12-21T17:24:21ZengMDPI AGRemote Sensing2072-42922018-07-01107110410.3390/rs10071104rs10071104Matching Multi-Sensor Remote Sensing Images via an Affinity TensorShiyu Chen0Xiuxiao Yuan1Wei Yuan2Jiqiang Niu3Feng Xu4Yong Zhang5School of Geographic Sciences, Xinyang Normal University, 237 Nanhu Road, Xinyang 464000, ChinaSchool of Remote Sensing and Information Engineering, Wuhan University, 129 Luoyu Road, Wuhan 430079, ChinaSchool of Remote Sensing and Information Engineering, Wuhan University, 129 Luoyu Road, Wuhan 430079, ChinaSchool of Geographic Sciences, Xinyang Normal University, 237 Nanhu Road, Xinyang 464000, ChinaSchool of Geographic Sciences, Xinyang Normal University, 237 Nanhu Road, Xinyang 464000, ChinaVisiontek Research, 6 Phoenix Avenue, Wuhan 430205, ChinaMatching multi-sensor remote sensing images is still a challenging task due to textural changes and non-linear intensity differences. In this paper, a novel matching method is proposed for multi-sensor remote sensing images. To establish feature correspondences, an affinity tensor is used to integrate geometric and radiometric information. The matching process consists of three steps. First, features from an accelerated segment test are extracted from both source and target images, and two complete graphs are constructed with their nodes representing these features. Then, the geometric and radiometric similarities of the feature points are represented by the three-order affinity tensor, and the initial feature correspondences are established by tensor power iteration. Finally, a tensor-based mismatch detection process is conducted to purify the initial matched points. The robustness and capability of the proposed method are tested with a variety of remote sensing images such as Ziyuan-3 backward, Ziyuan-3 nadir, Gaofen-1, Gaofen-2, unmanned aerial vehicle platform, and Jilin-1. The experiments show that the average matching recall is greater than 0.5, which outperforms state-of-the-art multi-sensor image-matching algorithms such as SIFT, SURF, NG-SIFT, OR-SIFT and GOM-SIFT.http://www.mdpi.com/2072-4292/10/7/1104image matchingmulti-sensor remote sensing imagegraph theoryaffinity tensormatching blunder detection |
| spellingShingle | Shiyu Chen Xiuxiao Yuan Wei Yuan Jiqiang Niu Feng Xu Yong Zhang Matching Multi-Sensor Remote Sensing Images via an Affinity Tensor Remote Sensing image matching multi-sensor remote sensing image graph theory affinity tensor matching blunder detection |
| title | Matching Multi-Sensor Remote Sensing Images via an Affinity Tensor |
| title_full | Matching Multi-Sensor Remote Sensing Images via an Affinity Tensor |
| title_fullStr | Matching Multi-Sensor Remote Sensing Images via an Affinity Tensor |
| title_full_unstemmed | Matching Multi-Sensor Remote Sensing Images via an Affinity Tensor |
| title_short | Matching Multi-Sensor Remote Sensing Images via an Affinity Tensor |
| title_sort | matching multi sensor remote sensing images via an affinity tensor |
| topic | image matching multi-sensor remote sensing image graph theory affinity tensor matching blunder detection |
| url | http://www.mdpi.com/2072-4292/10/7/1104 |
| work_keys_str_mv | AT shiyuchen matchingmultisensorremotesensingimagesviaanaffinitytensor AT xiuxiaoyuan matchingmultisensorremotesensingimagesviaanaffinitytensor AT weiyuan matchingmultisensorremotesensingimagesviaanaffinitytensor AT jiqiangniu matchingmultisensorremotesensingimagesviaanaffinitytensor AT fengxu matchingmultisensorremotesensingimagesviaanaffinitytensor AT yongzhang matchingmultisensorremotesensingimagesviaanaffinitytensor |