摘要
[目的]针对遥感图像舰船目标检测任务中轻量化和快速推理的需求,提出一种基于改进YOLOv5s的轻量化遥感舰船目标检测算法LR-YOLO。[方法]首先,主干网络采用ShuffleNet v2 Block堆叠方式,有效减少算法的参数量并提高计算速度;其次,设计区域选择模块Filter,选择感兴趣的区域,更充分地提取有效特征;最后,引入圆形光滑标签计算角度损失,对遥感舰船目标进行旋转检测,并采用可变形卷积,以此来适应几何形变,提升检测效果。[结果]在HRSC2016舰船数据集上的实验结果表明,该算法的检测精度达到92.90%,提高1.3%,并且算法参数量仅为基线模型的39.33%。[结论]该算法实现了轻量化和检测准确率的平衡,为轻量化遥感舰船目标检测提供了参考。
[Objective]This paper proposes a lightweight remote sensing ship target detection algorithm LRYOLO based on improved YOLOv5s to meet the lightweight and fast inference requirements of ship target detection tasks involving remote sensing images.[Methods]First,the backbone network adopts the ShuffleNet v2 block stacking method,effectively reducing the number of network model parameters and improving the computational speed;second,a region selection module filter is designed to select regions of interest and extract effective features more fully;finally,a circular smooth label is introduced to calculate angle loss and perform rotation detection on remote sensing ship targets,while deformable convolution is used to adapt to geometric deformation and improve detection performance.[Results]The experimental results on the HRSC2016 ship dataset show that the detection accuracy of the algorithm reaches 92.90%,an improvement of 1.3%,with the number of network model parameters only 39.33%that of the baseline model.[Conclusion]The proposed algorithm achieves a balance between lightweight and detection accuracy,providing references for remote sensing ship target detection.
作者
王浩臣
辛月兰
郭江
王庆庆
WANG Haochen;XIN Yuelan;GUO Jiang;WANG Qingqing(College of Physics and Electronic Information Engineering,Qinghai Normal University,Xining 810001,China)
出处
《中国舰船研究》
CSCD
北大核心
2024年第5期200-207,共8页
Chinese Journal of Ship Research
基金
国家自然科学基金资助项目(61662062)
青海省自然科学基金面上资助项目(2022-ZJ-929)。
