基于机器视觉的机器人避障算法研究
Research on Robot Obstacle Avoidance Algorithm Based on Machine Vision
蓝仁稳1
作者信息
- 1. 广东海洋大学电子与信息工程学院,广东 湛江 524088
- 折叠
摘要
为了解决机器人在自主导航中避障能力有限的问题,采用YOLOv5s目标检测模型来实现障碍物视觉感知.随后,将视觉感知算法获取的障碍物二维坐标和深度相机的深度值结合,形成实时导航信息.在避障方面,提出了自适应修正导航矢量场算法,以实现自主导航功能.通过在基于AirSim的仿真平台上的验证,结果显示避障成功率可达94%,表明该算法能够增强机器人的感知能力并提高其避碰性能.
Abstract
In order to solve the problem of limited obstacle avoidance ability in robot autonomous navigation,the YOLOv5s object detection model is adopted to achieve visual perception of obstacles.Subsequently,the two-dimensional coordinates of obstacles obtained by the visual perception algorithm are combined with the depth values of the depth camera to form real-time navigation information.In terms of obstacle avoidance,an adapted modified guidance vector field algorithm is proposed to achieve autonomous navigation function.Through validation on an AirSim-based simulation platform,the results show that the success rate of obstacle avoidance can reach 94%,indicating that the algorithm can enhance the robot's perception ability and improve the collision avoidance performance.
关键词
机器视觉/避障/AirSim/YOLOv5s/自适应修正导航矢量场Key words
machine vision/obstacle avoidance/AirSim/YOLOv5s/adapted modified guidance vector field引用本文复制引用
出版年
2024
NETL
NSTL
万方数据