光流结合特征提取的室内机器人避障技术研究
Research on Obstacle Avoidance Technology of Indoor Robots Based on Optical Flow and Feature Extraction
曹梦龙 1石梦鸽1
作者信息
- 1. 青岛科技大学自动化与电子工程学院,青岛 266042
- 折叠
摘要
为解决室内机器人正向准确避障问题,提出一种光流结合特征提取的视觉避障方法.将LK光流与多尺度思想结合,加入仿射变换,提高算法追踪角点的噪声鲁棒性,进而准确检测视场中的障碍物;通过评估障碍物风险程度,制定出碰撞时间(TTC)结合光流平衡策略的碰撞机制,引导机器人在无碰路径上移动.改进光流与特征提取结合能提高相机快速运动下的追踪成功率,制定的碰撞策略能够有效规避机器人前方的障碍物.实验表明,相机快速运动时,算法能提高追踪的准确性,有效检测出视场中的障碍物,引导机器人无碰撞行驶,具备较强的独立性和实时性.
Abstract
In order to solve the forward obstacle avoidance problem of indoor robots,a visual obstacle a-voidance method based on optical flow and feature extraction is proposed.LK optical flow is combined with multi-scale thought,and affine transformation is added to improve the noise robustness of tracking corner points,so as to accurately detect obstacles in the field of view.By assessing the risk degree of obstacles,a collision mechanism based on collision time(TTC)combined with an optical flow balancing strategy was developed to guide the robot to move on a collision-free path.The combination of improved optical flow and feature extraction can improve the tracking success rate of the camera under fast motion,and the collision strategy formulated can effectively avoid the obstacles in front of the robot.Experiments show that the algo-rithm can improve tracking accuracy,effectively detect obstacles in the field of view,and guide the robot to drive without collision when the camera moves quickly,with strong independence and real-time performance.
关键词
角点/光流/碰撞时间/平衡策略/避障Key words
corners/optical flow/time-to-collision/balancing strategy/obstacle avoidance引用本文复制引用
基金项目
山东省自然科学基金(ZR2020MF087)
出版年
2024
NETL
NSTL
万方数据