基于YOLOv5-SSA的水炮射击半遮挡目标检测
Semi-obstructed Object Detection in Water Cannon Shooting Based on YOLOv5-SSA
向河汉 1陈黎 1陈姚节1
作者信息
- 1. 武汉科技大学智能信息处理与实时工业系统湖北省重点实验室 武汉 430081
- 折叠
摘要
船舶水炮射击时容易产生水花,造成对射击目标的遮挡,现有水炮控制跟踪系统在水炮射击场景中容易丢失目标,难以获得有效的瞄准效果.论文提出一种基于深度学习的目标检测方法,将水炮射击时拍摄的图像作为输入,实时检测出要射击目标的位置,来达到良好的瞄准效果,实现对目标准确有效的射击.论文设计一个轻量级的深度学习目标检测网络YOLOv5-SSA对水炮射击的目标进行实时检测,在实际场景中进行测试时,在目标检测准确率达到93.2%的前提下,处理速度达到了21.18ms,实现了对目标的实时检测.
Abstract
When the ship's water cannon is shooting target,it is easy to produce splashes,which can block the shooting tar-get.The existing water cannon control and tracking system is easy to lose the target in the water cannon shooting scene,and it is dif-ficult to obtain an effective aiming effect.This paper proposes an object detection method based on deep learning,which takes the image taken during the shooting of the water cannon as input,and detects the position of the target to be shot in real time to correct the effect of object tracking and achieve accurate and effective shooting of the target.In this paper,a lightweight deep learning ob-ject detection network YOLOv5-SSA is designed to detect targets fired by water cannons in real-time.When tested in actual scenes,the processing speed reaches 21.18ms under the premise that the object detection precision reaches 93.2%,and it achieves real-time detection of the target.
关键词
目标检测/实时/深度学习/水炮Key words
object detection/real-time/deep learning/water cannon引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据