A scene matching method based on weighted Hausdorff distance combined with structure information

Lin Song,Cheng Yong-mei,Nan Liu,Song Chun-hua,Xu Ma
2013-01-01
Abstract:Traditional scene matching algorithms used for vision navigation are difficult to overcome the inference of noise, rotation and scale. A new method of scene matching is proposed, which is based on weighted Hausdorff distance combined with structure information. As structure information of feature points is introduced, the endpoints are extracted from the edge lines as the key points of the structure. The point matching cost of the shape context is computed to be as the weighted coefficient of the Hausdorff distance, which is used in scene matching for locating UAV. At last, matching simulation results demonstrate that the proposed method is effective and robust to noise, rotation and scale.
What problem does this paper attempt to address?