부산대학교 도서관

In the latest years great importance and research efforts have been put into safety and security aspects concerning the vehicular framework. In this connection, maritime surveillance is playing an important role, since situation awareness proves fundamental to ensure safety conditions at sea. In this work, we propose a novel maritime surveillance system based on Video Content Analysis, where vessel detection is performed automatically by a remote Machine Learning based target identification algorithm. As performance study, we carry out experimental tests analyzing the impact of packet loss, compression rate and transport protocol type on ship-to-ground communication over a satellite link, and their joint effects on video quality, transmission time and vessel detection accuracy.