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

作　　者：蓝仁稳 LAN Renwen(School of Electronic and Information Engineering,Guangdong Ocean University,Zhanjiang,Guangdong 524088,China)

机构地区：[1]广东海洋大学电子与信息工程学院,广东湛江524088

出　　处：《自动化应用》2024年第13期32-34,共3页Automation Application

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。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 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]