无人机平飞下激光雷达和双目视觉融合的SLAM建图

LIDAR AND STEREO CAMERA DATA FUSION IN UAV MOTION FOR SLAM MAPPING

针对无人机在较快运动下使用单一传感器难以有效构建地图的问题,提出一种基于LiDAR与双目融合的建图方法。利用LiDAR获取远处环境信息,提取物体的边界信息减少数据冗余,实现预先地图的构建;使用双目获取近处的信息,同时利用改进的特征金字塔法以实现特征快速提取,与LiDAR预建地图并行计算提高运行效率;最后将视觉特征与LiDAR预建地图融合构建二维地图。实验表明,该方法在无人机不同速度下可构建可靠性更高、更精确的环境地图,可为快速无人机的实时避障提供位置信息。

An unmanned aerial vehicle equipped with a sing sensor cannot construct an accurate map when UAV is moving at high speed.To handle this issue,a mapping method based on the fusion of LiDAR and stereo sensor is proposed.LiDAR was used to obtain the remote environment information,and the boundary of objects was extracted to reduce data redundancy,so as to realize the construction of the pre-map.The improved feature pyramid method was used to extract the near information of the UAV perceived by the stereo camera quickly,which improved the running efficiency with the parallel computing of LiDAR pre-map.The visual features and LiDAR pre-map were fused to build a 2D map.The experiments show that the proposed method can build more reliable and accurate environment map under different speed of UAV,which provides location information for fast UAV to avoid obstacles in real time.