On-Board Real-Time Ship Detection in HISEA-1 SAR Images Based on CFAR and Lightweight Deep Learning

Synthetic aperture radar (SAR) satellites produce large quantities of remote sensing images that are unaffected by weather conditions and, therefore, widely used in marine surveillance. However, because the hysteresis satellite-ground communication and massive quantity images, rapid analysis is not possible real-time information for emergency situations restricted. To solve this problem, paper proposes an on-board ship detection scheme based on traditional constant false alarm rate (CFAR) method lightweight deep learning. This can be SAR satellite computing platform to achieve near image processing data transmission. First, we use CFAR conduct initial then apply You Only Look Once version 4 (YOLOv4) obtain more accurate final results. We built a ground verification system assess feasibility our scheme. With help embedded Graphic Processing Unit (GPU) with high integration, achieved 85.9% precision experimental data, results showed time was nearly half required methods.