Ship detection and property extraction in radar images using hardware

In this work we investigate radar imaging satellites' dependency on ground stations to transfer the image data. Since synthetic aperture radar images are very big, only ground stations are equipped to process that much data in realtime. This is a problem for maritime surveillance as it creates delay between imaging and processing. We propose a hardware algorithm that can be used by a satellite to detect ships and extract information about them in real time, and since this information is smaller it can be relayed with significant reduction in delay. For ship detection, adaptive thresholding algorithm with exponential model is used. This algorithm was selected as it can be applied in real time. For the property calculation, a data accumulating, single-look, connected component labeling algorithm is proposed. This algorithm accumulates data about the connected components which is then used to calculate the properties of ships using image moments. The combined algorithm was then validated on RADARSAT-2 images using Matlab for software and co-simulation for hardware. The algorithm was able to detect ships and calculate the features with less than 5% error.