摘要
针对船舶在狭窄航道交汇时可能出现碰撞等问题,提出一种基于机器视觉的船舶检测与航行安全距离监测方法。这一方法以YOLOv5算法为基础,采用密集残差连接方式来提高浅层特征信息的复用率,加强特征传递,利用组合学习的方式提取船舶特征。在残差模块的输入端前加入空间注意力机制模块,对干扰特征进行过滤。对检测器进行解耦,提升图片的表征能力。通过YOLOv5算法、Sort算法,运用逆透视映射融合检测出船舶之间的距离是否达到碰撞阈值。这一方法可以为船舶行驶创造一个安全的环境。试验结果表明,这一方法检测的平均精度均值达到97.66%,检测精度高,能够有效预防船舶碰撞。
Aiming at problem such as collision that may occur when ships meet in narrow channel,a method for ship detection and monitoring of navigation safety distance based on machine vision was proposed.This method is based on the YOLOv5 algorithm,adopts the dense residual connection method to improve the reuse rate of shallow feature information,strengthens feature transfer,and uses combined learning to extract marine feature.A spatial attention mechanism module is added before the input of the residual module to filter the interference feature.Decouple the detector to improve image representation.Through YOLOv5 algorithm and Sort algorithm,the inverse perspective mapping fusion is used to detect whether the distance between ships reaches the collision threshold.This method can create a safe environment for ship to operate.The test result shows that the mean average precision of this method is 97.66%,and the detection accuracy is high,which can effectively prevent ship collision.
出处
《机械制造》
2022年第8期67-73,94,共8页
Machinery
基金
宁波市重大科技任务攻关项目(编号:2021Z100)
关键词
机器视觉
船舶
航行
安全性
Machine Vision
Ship
Navigation
Safety
