The paper presents new techniques for automatic segmentation, classification, and generic pose estimation of ships and boats in laser radar imagery. Segmentation, which primarily involves elimination of water reflections, is based on modeling surface waves and comparing the expected water reflection signature to the ladar intensity image. Shape classification matches a parametric shape representation of a generic ship hull with parameters extracted from the range image. The extracted parameter vector defines an instance of a geometric 3D model which can be registered with the range image for precision pose estimation. Results show that reliable automatic acquisition, aim point selection and real-time tracking of maritime targets is feasible even for erratic sensor and target motions, temporary occlusions, and evasive target maneuvers.
Segmentation, classification, and pose estimation of maritime targets in flash-ladar imagery
2012
2012
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Flash Ladar Flight Testing and Pathway to UAV Deployment
AIAA | 2010
|Robots and Vision: Super-resolution boosts flash-ladar accuracy
British Library Online Contents | 2010