Numéro : 2541 - Year : 2008
Automatic boat segmentation and enhancement for marina monitoring.
Habib DJEBALI, Laboratoires PROTEE
Valentin GIES, Laboratoires PROTEE
Michel GRIMALDI, Laboratoires PROTEE
Frédéric BOUCHARA, Laboratoire LSIS
International Autonomous Surface Symposium (ASSS 2008)
Docking and undocking boats are not currently identified in marinas outside the harbormaster’s office opening hours. This raises safety and economical issues. In this paper, we present a robust and automatic method for extracting boat shapes from video sequences containing both fixed elements (such as harbor installations) and moving ones (sea, birds…). The proposed algorithm can be split into 3 parts. First, a statistical real-time analysis of the video flow aims at extracting the passing boat sequences. Second, a multi-resolution boat tracking in the sequence allows an efficient characteristic image extraction of the considered scene. Third, a spatiotemporal and statistical filtering method relying on stochastic properties of the sea removes all irrelevant information on the scene. This analysis cuts out with accuracy the boat out of the video sequence, allowing to perform a supervised identification after. This work is part of an applied research project named SYREN involving both university laboratories (Protée – Université du Sud Toulon Var) and private companies (INSILIO, CS and MicroBe).
(Text in French only)
Advanced search
Year