Multi-Scale Object Detection Model for Autonomous Ship Navigation in Maritime Environment

Accurate detection of sea-surface objects is vital for the safe navigation autonomous ships. With continuous development artificial intelligence, electro-optical (EO) sensors such as video cameras are used to supplement marine radar improve that produce weak signals and small sizes. In this study, we propose an enhanced convolutional neural network (CNN) named VarifocalNet * improves object in harsh maritime environments. Specifically, feature representation learning ability model improved by using a deformable convolution module, redesigning loss function, introducing soft non-maximum suppression algorithm, incorporating multi-scale prediction methods. These strategies accuracy reliability our CNN-based results under complex sea conditions, turbulent waves, fog, water reflection. Experimental different conditions show method significantly outperforms similar methods (such SSD, YOLOv3, RetinaNet, Faster R-CNN, Cascade R-CNN) terms robustness objects. The obstacle were obtained imaging demonstrate performance model.