A new image matching algorithm based on multi-scale segmentation applied for UAV navigation

Image matching is an important UAV (Unmanned Aerial Vehicle) navigation method. Traditional area-based image matching methods have high accuracy but are sensitive to scale and rotation variations. In order to matching images which have geometric and illumination variations quickly and robustly, a new image matching algorithm based on multi-scale segmentation is proposed in this paper. In our method, we first extract stable image objects which have rich and reliable statistical information by object-oriented convexity model based multi-scale segmentation. Next, shape features like invariant moments and shape curves are extracted and used in local object matching. Afterwards, similar triangles criterion based global object matching which can match objects in spatial relationship is done. Finally, the centroids of matched objects are used as control points to register the real time image. Many experiments have proved that the proposed method is fast and robust.