Evaluating the velocity of ships from low-resolution SAR images

Synthetic aperture radar (SAR) is a fundamental tool in maritime surveillance for its weather and daylight insensitivity and large swath. Position, size and bearing of a target ship can be extracted with high accuracy from a SAR image. Additionally, those images contain several cues to estimate the target velocity. However, dealing with spatial resolutions of tens of meters, the applicability of some approaches is questionable. We formerly experimented two velocity estimation methods based on the features extracted from the wake of a sailing ship [1]: one is based on the wavelength of the cusp waves characterizing the Kelvin wake pattern; the other is based on the typical azimuth shift that affects moving targets imaged by a SAR system. Both methods directly return the radial component of the ship speed; the azimuthal speed, and thus the complete velocity vector, can finally be reconstructed using the estimated ship track. The mentioned techniques rely on the detection of significant parts of the wake in the SAR image. This is not always possible, especially with low-resolution images, due to a combination of unfavorable incidence angles and rough status of the sea.