A real-time on-board ship targets detection method for optical remote sensing satellite

Optical remote sensing satellite holds great potential for ship detection. However, it is challenging for real-time detection due to the relatively low resolution and complicated background. We propose a real-time on-board ship detection method based on statistical analysis and shape identification. First, Gaussian and median filter are used to reduce the periodical and pepper noise generated by the camera sensor system. Then, mathematical morphology processing is employed to remove the background interference and thus enhances the ship targets. Next, statistic analysis is performed on inspected and neighbor areas to distinguish the suspected ship target and pure-sea, land, islands or strong waves. Finally, features, such as the length-width ratio, circular degree of the suspected targets, are used to detect the targets. The proposed method was implemented on a single FPGA and was validated on real high-orbit optical satellite images. For an image as large as 1024×1024 of 8bits, the computation time was less than 10 seconds. The detection rate was above 90% while the false alarm rate was under 5%. The experimental results demonstrate the ability to support low-power consumption, miniaturization for the real-time ship targets detection on-board.