基于RGB-D的机器人跟随系统设计

Robot Following System Based on RGB-D

机器人跟随可广泛应用于物流、安防、家庭服务、医院等诸多领域,基于ROS平台设计了机器人跟随系统,包括机器人差速运动控制、深度图像处理和机器人跟随软件。在目标跟踪定位方面,提出了一种利用深度图像信息(RGB-D)的深度质心算法,该算法在采用KCF算法进行目标跟踪的基础上,计算跟踪矩形区域的深度质心值,然后剔除远高于深度质心值的背景区域,从而获得目标中心更加准确的位置。为验证算法,在不同距离对同一目标进行跟踪并对跟踪位置进行分析,结果表明深度质心算法的目标定位精度较KCF跟踪算法提升明显。

Robot following can be widely used in logistics,security,family services,hospitals and many other fields.In this paper,a robot following system is designed based on ROS platform,including robot differential motion control,image processing and robot following software.A new method is proposed to calculated the depth centroid of the RGB-D images for target locating and following.The method calculates the depth centroid value of KCF rectangular tracking region,and removes the background area which depth is far more depth centroid value,then it can get relative actual position of the target center where the robot will move to follow.In order to verify the method,we test it by tracking the same target at different distance positions.The results show that the depth centroid method can get more precise position than KCF in target locating and following.