3D Shape Mapping of Obstacle Using Stereo Vision Sensor on Quadrotor UAV
3D Shape Mapping of Obstacle Using Stereo Vision Sensor on Quadrotor UAV
复制标题
使用四旋翼无人机上的立体视觉传感器进行障碍物 3D 形状测绘
DOI：
10.2514/6.2014-0975
复制
发表时间：
2014
影响因子：
3.3
通讯作者：
Youdan Kim
中科院分区：
文献类型：
--
作者：
Jongho Park;Youdan Kim
针对四旋翼无人机，提出了基于立体视觉的未知障碍物形状映射算法。假设将立体视觉安装在四旋翼上以获得深度图。在通过深度图检测到障碍物后，通过从深度图中提取距离信息以及四旋翼的当前状态信息来映射障碍物的形状。由于立体视觉的固有特性，随着四旋翼向障碍物靠近，可以提取更精确的信息。在映射过程中，生成导航指令，使四旋翼在障碍物周围移动，从各个方向获取障碍物的信息，同时将障碍物保持在图像平面内。通过比较它们的计算效率差异来去除具有重叠信息的点。通过数值仿真验证了该映射算法的性能。
Shape mapping algorithm of unknown obstacle is proposed using stereo vision for a quadrotor unmanned aerial vehicle. Stereo vision is assumed to be mounted on the quadrotor to obtain a depth map. After detecting an obstacle through the depth map, mapping the shape of the obstacle is performed by extracting the distance information from the depth map along with the information of the quadrotor’s current state. Due to the inherent property of the stereo vision, more precise information can be extracted as the quadrotor moves closer to the obstacle. Guidance command is generated to make the quadrotor move around the obstacle to obtain the information of the obstacle from all directions while the obstacle is maintained inside the image plane during the mapping process. Points with overlapped information are removed by comparing their disparities for computational efficiency. Numerical simulation is performed to demonstrate the performance of the proposed mapping algorithm.