|
刘一博(1987—),男,博士,高级工程师,研究方向为水下无人航行器及集群系统。 |
Copy editor: 张培培
收稿日期: 2024-07-15
修回日期: 2024-08-02
网络出版日期: 2024-11-26
基金资助
*国家自然科学基金项目资助(U2141238)
Research on detection and tracking methods of unmanned ship water targets based on light vision
Received date: 2024-07-15
Revised date: 2024-08-02
Online published: 2024-11-26
针对水面无人船在复杂环境中的目标检测和跟踪问题,探索了基于光视觉的技术方法。利用改进的暗通道去雾方法和引导滤波进行图像预处理,以提高后续图像处理的准确性和效率。在目标检测方面,采用了YOLOv7算法,该算法通过对损失函数的优化,有效提高了目标检测的精度和召回率。为实现精准多目标跟踪,结合自训练模型权重和Sort算法,成功实施了目标持续追踪和中心点轨迹的精确标注。此外,通过无人船平台搭建双目相机系统进行目标测距,实验结果表明,该方法可以实现测距功能,平均相对误差为6.46%,这一成果不仅提升了无人船的导航与定位能力,还为水面安全监控提供了技术支持。研究结果显示,在水面无人船领域,通过融合先进的图像处理技术和机器学习算法,能够有效地解决目标检测和跟踪问题。
刘一博 , 邱昕雨 , 王天昊 , 高岩岫嵩 , 王银涛 . 基于光视觉的无人船水面目标检测与跟踪方法*[J]. 指挥控制与仿真, 2024 , 46(6) : 78 -86 . DOI: 10.3969/j.issn.1673-3819.2024.06.013
This study explores technical methods based on light vision to address the problem of target detection and tracking by surface unmanned ships in complex environments. We utilize an improved dark channel dehazing method and guided filtering for image preprocessing to improve the accuracy and efficiency of subsequent image processing. In terms of target detection, the YOLOv7 algorithm is used, which effectively improves the accuracy and recall rate of target detection by optimizing the loss function. In order to achieve accurate multi-target tracking, combined with self-trained model weights and Sort algorithm, continuous tracking of targets and accurate annotation of center point trajectories are successfully implemented. In addition, a binocular camera system is built on an unmanned ship platform for target ranging. Experimental results show that our method can achieve the ranging function with an average relative error of 6.46%. This result not only improves the navigation and positioning capabilities of unmanned ships, but also provides technical support for water surface safety monitoring. This research demonstrates that in the field of surface unmanned ships, target detection and tracking problems can be effectively solved by integrating advanced image processing technology and machine learning algorithms.
Key words: target detection; multi-target tracking; defogging
| [1] |
|
| [2] |
|
| [3] |
|
| [4] |
|
| [5] |
|
| [6] |
|
| [7] |
|
| [8] |
|
| [9] |
|
| [10] |
|
| [11] |
|
| [12] |
|
/
| 〈 |
|
〉 |