白永康, 刘清宇, 伍世虔. 基于kinect的室内移动机器人楼梯攀爬位姿快速估计方法[J]. 微电子学与计算机, 2021, 38(6): 45-52.
引用本文: 白永康, 刘清宇, 伍世虔. 基于kinect的室内移动机器人楼梯攀爬位姿快速估计方法[J]. 微电子学与计算机, 2021, 38(6): 45-52.
BAI Yong-kang, LIU Qing-yu, WU Shi-qian. Kinect-based fast estimation method of indoor mobile robot stair climbing pose[J]. Microelectronics & Computer, 2021, 38(6): 45-52.
Citation: BAI Yong-kang, LIU Qing-yu, WU Shi-qian. Kinect-based fast estimation method of indoor mobile robot stair climbing pose[J]. Microelectronics & Computer, 2021, 38(6): 45-52.

基于kinect的室内移动机器人楼梯攀爬位姿快速估计方法

Kinect-based fast estimation method of indoor mobile robot stair climbing pose

  • 摘要: 楼梯是室内移动机器人面对的主要障碍之一,攀爬过程中需要估计机器人相对于楼梯的位姿.大多数室内定位的方法均应用于二维空间平面,而攀爬楼梯属于三维空间的运动,多了高度上的位移,为位姿估计带来了很大的挑战. 本文提出了一种基于二维图像和部分深度信息的机器人攀爬楼梯时的位姿估计方法.该方法利用水平楼梯边沿线与楼梯台阶的高度和宽度边沿相交的特性,提出三线相交法来检测楼梯边沿点.然后结合边沿点的深度信息得到其在相机坐标系下的三维坐标,建立楼梯坐标系并求解位姿.同时,利用边沿点到相机距离的周期变化特性实现行走楼梯阶数的计数.最后利用kinect相机在楼梯上的移动模拟机器人的攀爬并对提出的方法进行测试,结果显示了其有效性和实用性.

     

    Abstract: Stairs are one of the main obstacles faced by indoor mobile robots. It is necessary to estimate the pose of the robot relative to the stairs during the climbing process. Most indoor positioning methods are applied to a two-dimensional space plane, and climbing stairs is a movement in three-dimensional space. Compared with the addition of height displacement in two-dimensional space, this brings great challenges to pose estimation. A pose estimation method for robots climbing stairs based on two-dimensional images and partial depth information is proposed in this paper. The method takes advantage of the characteristics of the intersection of the edge line of the horizontal staircase and the edge line where the height and width of the stair step are located, and proposes a three-line intersection method to detect the edge point of the staircase. Then combine the depth information of the edge point to get its three-dimensional coordinates in the camera coordinate system, and use the coordinates to establish the staircase coordinate system and solve the pose. At the same time, the periodic variation characteristic of the distance from the edge point to the camera is used to count the steps of walking stairs. Finally, the kinect camera is used to move on the stairs to simulate the climbing of the robot and the proposed method is tested. The results show its effectiveness and practicability.

     

/

返回文章
返回