Concatenating wide-parallax satellite orthoimages for simplified regional mapping via utilizing line-point consistency

Abstract

ABSTRACT Block adjustment is a common way for achieving large-scale mapping from several ortho-rectified satellite images derived from wide-parallax pushbroom (WPP) imaging. However, block adjustment’s accuracy can be affected by the amount of ground control points (GCPs), resolution of digital evaluation model (DEM) and unexpected and unpredictable imaging drift. In addition, block adjustment requires sensor model, which is hard to obtain and needs adjustment in real time. In view of these problems, this paper proposes a WPP blocks concatenation method for simplified regional mapping. First, considering distortions along different blocks, an adaptive-RPC model is proposed to ortho-rectified images. Then, a novel algorithm for stitching the ortho-rectified images is proposed, which consists of line-point consistency for matching neighbourhoods between two images, and a new energy function for suppressing artefact and distortions. Stitching experiments carried out on images from Orbita Satellite show that the stitching accuracy reaches 0.87 pixels, better than state-of-the-art methods, and absolute cumulative accuracy after stitching 10 orthoimages is only 3.99% higher than original one. Further simulation experiments with variable in satellite swing angles verify its robustness under complex scenes, while the estimation of cumulative error reiterates the necessity of performing orthorectification in prior. All results demonstrate that the proposed method can obtain accurate large regional mapping with distortions and artefact effectively suppressed.