Positioning a point target in an aerial image

This paper provides a novel method utilizing the recognized and verified support objects and the geometrical relationships to position a point target in an aerial image. The method benefits from requiring a small model database, representing objects simply, and shortening the time of both searching and matching the database data. Features of the support objects under translation, rotation and scale transformations as well as perspective projection are invariants. Moreover, a weighted LMedS algorithm is used to reduce noise influence and improve the accuracy of the positioning result. The experimental results demonstrate that the proposed method is effective and robust to position a point target in an aerial image.