Towards robust water region extraction for maritime surveillance

Although the concept of region of interest (ROI) is well known in video analysis, finding a suitable ROI has been hardly addressed in practical maritime surveillance such as for vessel detection and tracking. Videos from maritime surveillance cameras may contain irrelevant regions, such as shorelines, bridges and piers. As a result, non-relevant moving objects (e.g. cars on the shorelines) can be misleadingly detected by a vessel or ship surveillance system. This paper proposes a robust water region extraction method based on spatiotemporally-oriented energy features in combination with a mean shift clustering algorithm. The method targets not only the conventional RGB surveillance data, but also data from thermal cameras. Experimental results reveal that the proposed method performs water segmentation correctly for 93.67% of pixels on RGB and 94.7% of pixels on thermal sequences on the average, even in the presence of islands or other complex shoreline shapes.