FLNet: A Near-shore Ship Detection Method Based on Image Enhancement Technology

In the past few years, Synthetic Aperture Radar (SAR) has been widely used to detect marine ships due its ability work in various weather conditions. However, imaging mechanism of SAR, there is a lot background information and noise similar images, which seriously affects performance ship detection models. To solve above problems, this paper proposes new model called Feature enhancement Land burial Net (FLNet), blends traditional image processing methods with object approaches based on deep learning. We first design SAR threshold segmentation method, Salient Otsu (S-Otsu), according difference between background. better eliminate we further combine such as Lee filtering. These constitute Enhancement Module (FEM) that mitigates impact data overall model. alleviate influence land detection, Burial (LBM) morphological differences areas. Finally, these two modules are added You Only Look Once V5 (YOLO V5) form our FLNet. Experimental results Ship Detection Dataset (SSDD) dataset show FLNet comparison YOLO accuracy when performing improved by 7% recall rate 6.5%.