LFSA‐YOLOv7: A Novel Method for Ship Detection in Remote Sensing Images with Complex Backgrounds

Abstract

Ship target detection is a critical component for marine environmental monitoring and the protection of maritime rights and interests. However, existing methods for ship detection in remote sensing imagery often face challenges such as false positives and missed detections, particularly in the presence of complex backgrounds and multi-scale targets. To address these issues, we propose a novel method called LFSA‐YOLOv7, (LDConv {Linear Deformable Convolution}; SPPFCSPC‐G {Spatial Pyramid Pooling Fast Cross Stage Partial Channel ‐ Group}, SimAM{Simple Attention Module}, and Alpha‐CoIoU {Alpha‐Coordinate Point Orienta tion ‐ Intersection over Union}) which is designed to enhance the accuracy and robustness of ship detection in complex backgrounds: (a) a novel network architecture (LDConv + SPPFCSPC‐G + SimAM) to enhance the model’s ship detection capabilities, and (b) a novel Alpha‐CoIoU loss function to improve the model’s ship localization accuracy. To evaluate the performance of our algorithm in complex backgrounds, we have developed a novel ship classification detection (SCD) data set and conducted comprehensive experiments. Furthermore, we have validated the generalization of our algorithm across various remote sensing ship data sets. The experimental results demonstrate that our algorithm outperforms previous techniques, exhibiting strong generalization and robustness, making it effectively suitable for ship detection in complex backgrounds. The SCD data set and code will be available at https://github.com/wionn/LFSA-YOLOv7.git.