On Payload Spatial and Spectral Resolutions for Automatic Ship Detection in Satellite Images

Maritime situational awareness can be enhanced using imagery provided by low earth orbit (LEO) satellites which have the capability to regularly monitor coastal zones. Successful ship detection relies on both payload imaging performance and smart onboard data processing techniques. In this study, a multistage ship detection algorithm is proposed, where the spectral, spatial and geometric properties of suspected ships are exploited for effective maritime target mapping. Multispectral images produced by the Landsat OLI and the Sentinel 2 MSI sensors of four locations were used in this study and they present a variety of scenarios that a ship detection algorithm could face. We observed that the NIR and SWIR bands produced a higher number of correctly detected ships when compared to the visible spectral bands and that the spatial resolution had a more significant impact than spectral resolution on true target detection levels.