摘要
针对航母飞行甲板目标检测背景复杂,且牵引设备以及作业人员等小目标检测效果有限等问题,提出了一种改进SSD算法的航母舰面多尺度目标检测算法。通过融合网络6个不同尺度的特征层,以及调整了适应于多尺度目标检测的网络先验框的大小,增强了网络对多尺度目标的检测性能。通过消融实验对方案进一步优化,改进后算法对于不同尺度目标检测的平均准确率的均值mAP提高了2.74%,对于牵引车和人员等小目标检测准确率分别提升了4.0%和5.1%。在与同类检测算法的性能对比中,论文算法检测准确性和实时性均处于较高水平。
In view of the complex background of aircraft carrier flight deck target detection,and the poor detection effect of small targets such as traction equipment and operators,an improved SSD algorithm is proposed for the aircraft carrier surface multi-scale object detection algorithm. By fusing the feature layers of the network with six different scales,and adjusting the size of the network a priori frame suitable for multi-scale object detection,the algorithm’s detection performance of different scale object on the ship surface is improved,especially the detection ability of small object. Experiments show that the improved algorithm improves the average mAP of the average accuracy of object detection at different scales by 2.74%,and increases the accuracy of detection of small objects such as tractors and personnel by 4.0% and 5.1%,the solution is further optimized through ablation experiments. In comparison with similar detection algorithms,the detection accuracy and real-time performance of the algorithm in this paper are both at a high level.
作者
朱兴动
田少兵
范加利
王正
ZHU Xingdong;TIAN Shaobing;FAN Jiali;WANG Zheng(Coast Guard Academy,Naval Aviation University,Yantai 264001;Department of Ship Surface Aviation Support and Station Management,Naval Aviation University(Qingdao Campus),Qingdao 266041)
出处
《舰船电子工程》
2022年第2期42-47,共6页
Ship Electronic Engineering
基金
军队科研基金项目“航母舰面舰载机牵引作业辅助系统关键技术研究”资助。
关键词
目标检测
深度学习
特征融合
数据增强
多尺度目标
object detection
deep learning
feature fusion
data enhancement
multi-scale object
