摘要
为了确保水上应急救援的及时性和可靠性,提高水上应急救援装备的智能化水平,提出一种水上应急救援无人船系统。首先提出落水人员识别模块,基于深度学习的单阶段检测算法,选择YOLOv5作为救援目标检测的网络模型,能够实现目标精准检测;以针孔相机模型的定位算法,定位落水人员的位置;在此基础上研究自主导航模块,通过LOS制导算法,实现应急救援无人船的自主导航和避障,通过轨迹跟踪实验,达到较好的跟踪效果;最后设计的自动救援模块,当距离目标点1 m范围内由救生系统抛掷救生圈实现救援。实验结果表明,能够满足江河湖泊应急救援的需求,提高救援行动的成功率。
In order to ensure the timeliness and reliability of water emergency rescue and improve the intelligent level of water emergency rescue equipment,an unmanned ship system for water emergency rescue is proposed.Firstly,the identification module of people who fell into the water is proposed,which is based on the single-stage detection algorithm of deep learning,and YOLOv5 is selected as the network model of rescue target detection,which can achieve accurate target detection,and the location algorithm of pinhole camera model is used to locate the location of people who fall into the water.On this basis,the autonomous navigation module is studied,and the autonomous navigation and obstacle avoidance of the emergency rescue unmanned ship are realized through the LOS guidance algorithm,and a better tracking effect is achieved through the trajectory tracking experiment.Finally,the automatic rescue module is designed,when the lifebuoy is thrown by the life-saving system within 1 meter from the target point.The experimental results show that it can meet the needs of emergency rescue of rivers and lakes and improve the success rate of rescue operations.
出处
《科技创新与应用》
2024年第4期50-54,共5页
Technology Innovation and Application
基金
河北省科技厅重点研发科技计划项目(21375412D)。
关键词
应急救援
无人船
落水人员识别
深度学习
LOS制导
emergency rescue
unmanned ship
identification of people who fall into the water
deep learning
LOS guidance
