Research on Sea Surface Ship Identification Based on Improved YOLOv10n Network

Absrtact-In order to solve the problems of insufficie nt recognition accuracy of the existing target detection models in $\mathbf{t}$ he identification and classification task of marine ships, an impro ved model of marine ship type recognition based on YOLOv10n w as proposed, and by comparing the classical models such as YOL Ov5n, YOLOv6n, OLOv8n, and YOLOv9n, it was found that YO LOv10n had the highest detection accuracy. The deformable conv olution module is introduced, which effectively improves the detec tion accuracy of the model. The CBAM attention mechanism is in troduced into the backbone network to enhance the attention of t he model to important features. Compared with the original YOL Ov10n, the enhanced YOLOv10n-v2 algorithm has a 0.7% impro vement in [email protected]. Compared with the original YOLOv10n, th e accuracy of the YOLOv10n-v2CB algorithm with CBAM attenti on mechanism is increased by 3.4%, and the accuracy of the algor ithm is increased by [email protected] to 96.9%. The improved YOLOv1 0n-v2CB algorithm significantly improves the detection ability of surface ships.