Target identification in complex maritime scenes

There are a growing number of civilian and military applications where it is necessary to obtain identification and tracking information of marine vessels that are manoeuvring in a complex maritime scene. The driving force behind this research is the need for surveillance and safety, particularly within harbours, where many maritime incidents and accidents occur. The research work developed here identifies and tracks maritime vessels moving within the narrow confines of a harbour entrance. Using a number of consecutive image frames, a hierarchical statistical motion pyramid is constructed, this is used to extract motion at different levels within each image. To identify and track the vessels as they move in the image a high-level reasoning process is used. This process associates the extracted motion dynamics at the different levels within the pyramid with the vessels concerned. This paper describes algorithms which use a visual based hierarchical motion scheme to identify and track multiple maritime vessels, typically boats and ships moving within a harbour scene