A Cognitive SAR Concept for Ship Detection using Support Vector Machines

Abstract

Cognitive radar is a new acquisition technique that forms a closed loop between radar receiver, radar transmitter and environment, similar to the perception-action cycle in human cognition. The continuous adaptation of the acquisition parameters based on previously acquired information also harbours great potential for future SAR missions. As an example, this paper presents a spaceborne cognitive SAR concept for ship detection. The concept foresees a two-stage process to improve the overall ship detection probability compared to conventional approaches. First, a wide-swath mode with coarse resolution is utilized to cover a large maritime area. From these SAR data, the positions of potential ships shall be detected, however, due to the low signal-to-clutter ratio, with a high false alarm rate. In the second step, a high-gain mode with fine resolution is used to look at the presumed ship positions and either confirm or reject the presence of ships with high fidelity. This radar concept could be realized on a single platform using a hybrid mode. In the context of this investigation, the cognitive functionality is distributed on two separate SAR satellites operated in a convoy configuration, where the leading satellite performs the first coarse-detection step and the companion satellite implements the high-fidelity detection step including intelligent digital beamforming of one or more spotlight beams accessible via phased array antennas.