Aiming at the problem of ship hull number recognition, this paper proposes a real-time ship's hull number recognition method for unmanned surface vehicles (USVs).

Based on a one-stage object detection model (e.g. YOLO), the attention mechanism is introduced to make the network more sensitive to the target area by the spatial information interaction module and divided attention method. Considering the effect of prior knowledge on accuracy, the adaptive anchor method and positive sample assignment strategy are utilized to improve the accuracy of regression. Aiming to resolve the problem of slow convergence at the beginning, the loss function is redesigned to speed up the convergence and enhance the stability of the network in the training phase. Finally, the proposed method is deployed in a USV to validate the availability of the recognition performance.

The results shows that the proposed method can achieve the recognition of ships and hull numbers simultaneously under Sea State 3 conditions, and has a 14% improvement in mean average precision (mAP) compared with the original model, with the ability to perform recognition in real time.

The results of this study indicate that the proposed method can be applied to USVs to perform hull number recognition, even under complex ocean conditions.