Shipboard aircraft autonomic landing is a very challenging task in the research of unmanned combat platform. In one image, the bounding boxes as an object in most successful object detectors, which is not only wasteful, inefficient, and requires additional postprocessing, but cannot meet the requirements of deck surface (small object) detection in large scene. In this paper, the CenterNet method, is simpler and faster than the object representation methods of past studies. The detector uses key point estimation to find deck surface center point and regresses to all other pose estimation parameters, such as depth, 3D sizes and orientation. CenterNet achieves the best results on our simulated deck surface dataset. And the same approach is used to estimation 3D bounding box of deck surface for point-based pose estimation. CenterNet is suitable for shipboard aircraft autonomic landing and runs in real-time.
Application of CenterNet Technology in the Process of Shipborne Aircraft Autonomous Landing
2022-04-08
463371 byte
Conference paper
Electronic Resource
English
Ear Detection based on CenterNet
IEEE | 2020
|Shipborne Helicopter - Landing, Security, Manoeuvring, and Traversing: Options and Implications
British Library Conference Proceedings | 1997
|AUTONOMOUS LANDING SYSTEMS AND METHODS FOR VERTICAL LANDING AIRCRAFT
European Patent Office | 2022
|