An automatic airborne image mosaicing method based on the SIFT feature matching

Image mosaicing is one of the core technologies in airborne remote sensing system and plays a very important role in generating the panorama of several aerial images. This paper presents a method for aerial image mosaicing based on the Scale Invariant Feature Transform (SIFT) algorithm. Firstly, a pre detection for feature points is operated by using Harris detector. The SIFT computing for feature points´ scale, position and direction is performed next and a rough matching is followed. Then Mahalanobis distances of the feature points are considered to improve matching precision, finally the Least Square (LS) method is adopted for mosaicing two images. Experiment results show this method performs well in aerial image mosaicing and the physical error is reduced to nearly 1 meter on average.