Multi-Algorithm Evaluation of Ship Detection Accuracy Using Synthetic Aperture Radar for the Radarsat Constellation Mission
This study investigates a 2-stage ship detection algorithm for the SAR Radarsat Constellation Mission. The three beam modes on the RCM with the most potential for ship detection are tested using four different dual polarizations: two linear and two compact polarizations. The first stage is an intensity-based detection; five different implementations of the detection are investigated, using a combination of the likelihood ratio test and constant false alarm rate algorithms. The second stage of the detection algorithm is a decomposition stage which estimates the proportion of single-bounce, double-bounce and volume backscatter from each pixel and identifies ship pixels based on their scattering mechanisms. The combination of the intensity-based and decomposition algorithms fails to identify and remove sidelobe detections from the scenes. Therefore, a study of three possible sidelobe removal algorithms is undertaken to attempt to remove these false detections from the results. The best option identified is a refined decomposition threshold, developed in this work, to use the volume scattering information to improve the removal of sidelobe pixels. The end results are promising with the majority of extra detections removed from the scenes for all three RCM beam modes tested and fewer than one verified target per scene missed by the detections. The compact polarizations outperformed the linear polarizations for both the intensity-based and decomposition algorithms.
Radarsat Constellation Mission, SAR, Ship Detection
Mantey, V. D. (2018). Multi-Algorithm Evaluation of Ship Detection Accuracy Using Synthetic Aperture Radar for the Radarsat Constellation Mission (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/31762