A new approach for geometric correction of UAV-based pushbroom images through the processing of simultaneously acquired frame images

Abstract

Video-assisted navigation is a strategy that hyperspectral camera manufacturers usually use to deal with the geometric distortions imposed on raw pushbroom images, due to perturbations. In this strategy, a simultaneous video is captured during a pushbroom raw image acquisition. This article proposes an automatic geometric correction method applicable to the sensors utilizing this strategy that does not need ground control data. This method starts with an inter-calibration procedure between the two sensors, which allows using frame-based images in the correction procedure. Then, the method uses sequential transformations among consecutive video frames to produce geometrically corrected products. The experimental results demonstrate that the proposed method's sequential nature makes it so flexible that it causes more reduction in the scenes' random geometric distortions than the common methods. Hence, the accuracy of 2D and 3D transformations relating the image and the ground spaces is increased by 66.9% compared to the raw pushbroom images.