基于视觉优化的移动机器人位姿检测方法研究  

Research on pose detection of mobile robot based on visual optimization

作　　者：潘涛[1] 王翔宇[2] 武权伟 卫文超 Pan Tao;Wang Xiangyu;Wu Quanwei;Wei Wenchao(Suzhou Joint Graduate School,Southeast University,Jiangsu Suzhou,215123,China;School of Automation,Southeast University,Jiangsu Nanjing,210096,China)

机构地区：[1]东南大学苏州联合研究生院,江苏苏州215123 [2]东南大学自动化学院,江苏南京210096

出　　处：《机械设计与制造工程》2025年第3期67-72,共6页Machine Design and Manufacturing Engineering

基　　金：国家自然科学基金(61873060)。

摘　　要：针对传统方法在移动机器人位姿检测过程中存在的目标丢失和数据漂移等问题,提出了一种基于ViT深度学习框架的视觉跟踪器,优化了利用ArUco码作为辅助特征标识的位姿检测方法,有效改善了检测效果。通过ROS系统分布式通信方法构建视觉处理平台和移动机器人之间桥梁,测试了不同数据集下跟踪器的性能和跟踪器的加入对整个位姿检测方法的提升效果,实验结果验证了提出的基于视觉优化的移动机器人位姿检测方法的可行性,可以较好地支撑移动机器人展开实时任务。A visual tracker is proposed based on ViT framework and ArUco markers are applied as auxiliary feature identifiers for pose detection of mobile robot,which effectively addresses the issues of target loss and data drift in mobile robot pose detection and improves detection performance.A visual processing platform is constructed in ROS environment which can build communication between the visual tracker and mobile robots via distributed communication mechanism.Performance of the tracker is tested on different datasets and the overall improvement of the pose detection method is evaluated by incorporating the visual tracker.The experimental results confirm the feasibility of the proposed visual optimization-based mobile robot pose detection method,which can effectively support mobile robots in real-time tasks.

关 键 词：目标跟踪 自注意力机制 ROS 位姿检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]