Automatic detection in a maritime environment using superresolution images

In the context of naval surveillance, shipboard Electro-Optical (EO) sensor systems can contribute to the detection, classification and identification of surface objects. Focusing on the detection process, our previous research offers a method using low-order polynomials for background estimation, which can be used for the automatic detection of objects in a naval environment. The polynomials are fitted to the intensity values in the image after which the deviation between the fitted intensity values and the measured intensity values are used for detection. The research presented in this paper, focuses on the impact of super-resolution algorithms on this detection process. Images enhanced by SR algorithms are expected to be mainly beneficial for classification purposes, regardless whether the classification is automatic or operator driven. This paper analyses the influence of SR algorithms on the detection performance in relation to the increase of computational complexity. The performance of the detection approach is tested on extensive dataset of maritime pictures in the Mediterranean Sea and in the North Sea collected on board of a frigate. We have found that for a good super-resolution image in this environment the sensor should be stabilised while recording and, for fast or near objects or when recording in heavier weather, should have a high frame rate and/or low exposure times.