No abstract

No abstract

A novel image registration algorithm was presented based on features point and clustering. Firstly, the believable matching points could be extracted from reference image and pre-registration image by using from coarse to fine matching, and which could be used to carry out geometric transform parameters. Meanwhile the map image was obtained by transform pre-registration image to reference image in the space coordinate according to calculated transformation parameters. Secondly, the believable matching points could be extracted from reference image and map image by using from coarse to fine matching. In addition, the anomaly detection algorithm based on cluster could eliminate the maximum error mismatching point. As a result, the translation matrix between reference image and map image could be obtained. Finally, image registration could be come out according to translation matrix correcting. Compared with traditional methods, the results of experiment with indoor and outdoor test images shown the proposed algorithm was effective and could obtain high precision registration image.

Visual feature extraction is foundation of lunar surface topography reconstruction. The paper studies visual feature extraction for lunar surface images, and then offers the visual feature extraction algorithm for lunar surface images. The experimental results show that the algorithm has good adaptability to illumination change, scale change and rotation transformation in the lunar unstructured environment. Meanwhile, the paper can extract the more accurate visual features of lunar surface images.

No abstract