Improved YOLOv5 with Attention Mechanism for SAR Ship Target Detection in Complex Environment

Jiajia Feng ； Zhijing Xu ； Zhi Liu

《International Core Journal of Engineering》 8卷6期 (2022/06) Pp. 290-301

https://doi.org/10.6919/ICJE.202206_8(6).0041

摘要

For ship target detection in synthetic aperture radar (SAR) images in complex environments, an algorithm based on improved YOLOv5 is proposed. First, a new feature pyramid network is proposed by using residual network thinking, which adds weighted cross-scale connections and reduces the loss of small target feature information; Secondly, the CBAM attention module suppresses the learning of background features and improves the feature learning ability of ship targets in complex backgrounds; Finally, the Soft-NMS linear penalty function is introduced to reduce the loss of the bounding box of the ship in the complex environment and improve the average accuracy. The experimental results show that the method proposed in this paper can effectively detect the ship target in the complex background, the detection speed is fast, and the missed detection rate is low.

關鍵字

SAR ； Ship Detection ； YOLOv5 ； Feature Pyramid Network ； Attention Mechanism

參考文獻

Zhang Y H , Ma H T , Yu Z X . Application of the method for prediction of the failure location and time based on monitoring of a slope using synthetic aperture radar[J]. Environmental Earth Sciences, 2021, 80(21).

Pappas O , Achim A , Bull D . Superpixel-Level CFAR Detectors for Ship Detection in SAR Imagery[J]. IEEE Geoscience and Remote Sensing Letters, 2018.

Cao D , Chen Z , Gao L. An improved object detection algorithm based on multi-scaled and deformable convolutional neural networks[J]. Human-centric Computing and Information Sciences, 2020, 10(1):14.

Girshick R , Donahue J , Darrell T , et al. Region-Based Convolutional Networks for Accurate Object Detection and Segmentation[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2015, 38(1):142-158.

Girshick R. Fast R-CNN[J]. Computer Science, 2015.