Ship Detection Algorithm for SAR Images Based on Lightweight Convolutional Network

Although ship detectors in synthetic aperture radar (SAR) images have continuously advanced the state-of-the-art performance in recent years. It is still difficult to balance the accuracy and efficiency. In this paper, we propose a ship detection algorithm for SAR images based on lightweight convolutional network. First, the Top-hat layer is designed by introducing the Top-hat operator, and Region Proposal Network (RPN) is constructed based on the layer to conduct rapid screening of SAR ship candidate regions. Second, the Facebook Berkeley Nets (FBNet) is introduced to accurately locate the SAR ship target in the candidate region and the Differential Neural Architecture Search technology is used to optimize the parameters of the network structure. Finally, the proposed ship detection framework is validated on the SAR ship datasets with other methods.