Affine ＆ Scale-Invariant Heterogeneous Pyramid Features for Automatic Matching of High Resolution Pushbroom Imagery from Chang＇e 2 Satellite

• Published in: 地球科学学刊：英文版 2016 (4), p.716-726
• Main Author: Sheng Yang Shibing Zhu Zhenju Li Xuejun Li Tao Liu Jue Wang Jianwei Xie
• Format: Article
• Language: English
• Subjects: 仿射变换; 卫星; 异构; 特征提取; 自动匹配; 航空影像; 金字塔; 高分辨率
• ISSN: 1674-487X; 1867-111X

The automatic feature extracting and matching for large amount of linear pushbroom imagery with higher and higher resolution is urgent and challenging in three dimensional reconstructions, remote sensing and mapping. Affine ＆ scale-invariant heterogeneous pyramid feature is proposed in this paper, along with the new scale-invariant analysis method, the detecting of the key points, the affine ＆ scale-invariant descriptor, the steering method of the matching, and the quasi-dense matching algorithm based on the extensive epipolar geometry. The automatic matching is devised for the linear pushbroom imagery. The whole process is executed on lunar images of the highest resolution of ~7 m/pixel（or ~1 m/pixel in the lower orbits） from the Chinese Chang＇e 2 satellite, it runs robustly at present, and resulting in large amounts of well-distributed-correspondences with accuracy of 0.3 pixels and excellent reliability, which gives great support for the further exploration both on the Moon and the Earth.