Neighborhood geometry based feature matching for geostationary satellite remote sensing image

In this paper, we focus on Global Self-consistent, Hierarchical, High-resolution Geography (GSHHG) database registration for remote sensing images taken from geostationary meteorological satellites. While the accuracy of feature matching is the key component. To improve it, we propose a neighborhood geometry-based feature matching scheme which includes three steps: neighborhood coding, verification and fitting. (1) Neighborhood coding represents landmarks of GSHHG as a descriptive bit-matrix, and quantifies remote sensing images to a probability-based edge map and a binary geometry-based edge map. As a result, both gradient and geometry similarity of local features in the remote sensing image and GSHHG can be measured. (2) Neighborhood verification is to encode spatial relationship among local features in neighbor, and discover outliers. (3) Neighborhood fitting fits the shorelines of GSHHG with the landmarks registered by neighborhood verification to improve recall. Experimental results on 25 pairs of newly annotated images show that the proposed method is competitive to several prior arts with respect to matching accuracy. What is more, our method is significantly more efficient than others.