Accuracy improvement of Best Scanline Search Algorithms for Object to Image Transformation of Linear Pushbroom Imagery

Unlike the frame type images, back-projection of ground points onto the 2D image space is not a straightforward process for the linear pushbroom imagery. In this type of images, best scanline search problem complicates image processing using Collinearity equation from computational point of view in order to achieve reliable exterior orientation parameters. In recent years, new best scanline search methods have been proposed to use in respective applications. This article tries to improve two previous methods named image space sequential search and bisecting window search. The proposed approach has a three-stage procedure. At first, considering exterior orientation parameters of reference line, approximate time for ground point is estimated with a high precision using sequential search and bisecting window search. In the second stage, the neighboring scanlines of the ground point are determined in image space. Finally, in the third stage the best scanline is calculated with a subpixel precision by interpolating the neighboring scanlines. Experimental results from two different datasets show that the proposed method outperforms the two foresaid existing methods which results in 13% accuracy improvement using real ground points and 30% accuracy improvement using simulated ground points.