Uçak ve iha platformlarından elde edilen eğik görüntülerin nokta tabanlı eşleştirilmesi

Images acquired from airborne oblique sensors and from Unmanned Aerial Vehicles (UAV) are widely used. On one hand, oblique imaging allows better visualization of side views of objects in three dimensional spaces. On the other hand, UAV imaging closes the gap between aerial and terrestrial photogrammetry. Therefore, combining airborne oblique and UAV images with a precise feature matching strategy can provide reliable information that can be useful for applications such as dense point cloud extraction for 3D city modeling, visualization, textured 3D city models, and so forth. In this study, a novel framework for the point-based feature matching of the airborne oblique and UAV imagery is presented. The proposed framework makes use of the powerful A-KAZE descriptor for feature extraction in both oblique and UAV images. Feature extraction with an iterative scheme is developed to construct tentative matches as many as possible. During the iterations, Brute Force matching is utilized for the initial matching of the corresponding features and left-right consistency check together with Lowe's nearest-next distance ratio test are forced to filtering erroneous matches. In order to extract putative matches from the tentative matches, three different strategies are implemented. Each strategy employed outlines a different robust method for the selection of matching points along with the epipolar constraint enforced between the two datasets. The developed framework is tested for image pairs acquired over the DortmundCentre, Germany, from the International Society for Photogrammetry and Remote Sensing (ISPRS) image orientation benchmark dataset. The proposed framework yields successful results in terms of matching precision and provides a nice balance between the true-positive and false-positive matches. Besides, the results of the proposed framework for two different test pairs outperformed the results of the previously developed approaches in the literature.