Landsat remote sensing images registration using the center points of extracted lakes as ground control points in pan-arctic area

Abstract

Image registration is the basis of lakes change detection from large volumes of remotely sensed images. Since Alaska area of the United State is covered by ice, snow and lakes all the whole year, it is difficult to find some valid Ground Control Points (GCPs) from the images, which adds the difficulty of image registration for lakes change detection and analysis. On the basis of lakes change regulation, we propose that the deepest point of a lake is the most stable point with the shape change of the lake, and these points can be used as GCP Points for image registration. The center point of a polygon can be computed by the Voronoi diagram of the polygon. After analyzing the Voronoi diagram computing procedure of a simple polygon, this paper presents the Voronoi generation method of a complex polygon, and then gives its algorithm complexity analysis. In the end of this paper we applied the algorithm to the largest inner circle seeking of many lakes in Alaska, and the results show that the proposed algorithm can compute the center points of all the largest inner circle of the lakes with a high efficiency and can gain a prefect registration effect based on these points.