基于改进RetinaFace和YOLOv4的船舶驾驶员吸烟和打电话行为检测  被引量:4

Smoking and calling behavior detection of ship officers based on improved RetinaFace and YOLOv4

在线阅读下载全文

作  者:王鹏[1] 尹勇[1] 宋策 WANG Peng;YIN Yong;SONG Ce(Key Laboratory of Marine Simulation&Control for Ministry of Transportation,Dalian Maritime University,Dalian 116026,Liaoning,China)

机构地区:[1]大连海事大学航海动态仿真和控制交通行业重点实验室,辽宁大连116026

出  处:《上海海事大学学报》2022年第4期44-50,共7页Journal of Shanghai Maritime University

基  金:国家重点研发计划(2019YFB1600602)。

摘  要:针对船舶驾驶员值班过程中吸烟和打电话行为造成注意力分散威胁船舶航行安全的问题,提出一种基于改进RetinaFace和YOLOv4的吸烟和打电话行为检测算法。采用改进的RetinaFace网络提取人脸感兴趣区域,使用改进的YOLOv4目标检测模型检测该区域内是否存在香烟或手机,从而识别船舶驾驶员的吸烟和打电话行为。实验结果表明,本文算法具有较高的检测精度和速度,在自建数据集上的类平均精度(mean average precision,MAP)高达98.51%,误检率仅为3.2%。使用PyQt开发图形界面程序。该算法可以准确识别出驾驶员的吸烟和打电话行为,能够较好地适应船舶驾驶台的复杂环境,满足实时检测的要求。Aiming at the problem of distracted attention caused by smoking and calling behaviors of ship officers on watch and threatening the safety of ship navigation,an algorithm for smoking and calling behavior detection based on the improved RetinaFace and YOLOv4 is proposed.The improved RetinaFace network is used to extract the regions of interest of faces,and the improved YOLOv4 object detection model is used to detect whether there are cigarettes or mobile phones in the area,so as to identify ship officers’smoking and calling behaviors.Experimental results show that the algorithm is of higher detection accuracy and speed.The mean average precision(MAP)on the self-built dataset is as high as 98.51%,and the false detection rate is only 3.2%.The graphical interface program is developed by PyQt.The algorithm can accurately identify the ship officers’smoking and calling behaviors,adapt to the complex environment of the ship bridge better,and meet the requirement of real-time detection.

关 键 词:吸烟行为检测 打电话行为检测 RetinaFace YOLOv4 数据增强 

分 类 号:U676.1[交通运输工程—船舶及航道工程]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象