基于RGB-D数据的SLAM算法

SLAM algorithm research based on RGB-D data

本文提出了一种基于RGB-D数据的SLAM算法,通过对相机获取的图像进行ORB特征点的提取与匹配,估计相机运动关系,实现点云拼接,最后会得到全局一致的点云地图和轨迹。为了消除误差积累引起的干扰,引入通用图优化库g2o,得到光滑的优化轨迹;在后端回环检测的过程中,引入关键帧选取机制,提高点云地图的生成效率,减少其消耗的存储空间。实验结果表明,本文的研究方法在RGB-D SLAM算法中具有可行性,并且能够满足实时性要求,具备较高的精度。

This paper proposes a SLAM algorithm based on RGB-D data.By extracting and matching the ORB keypoints of the images obtained by the camera,the camera motion relationship is estimated and the point clouds are stitched.Finally,a globally consistent point cloud map and trajectory are obtained.In the process of detection of the backend end loop closure,a key frame selection mechanism is introduced to improve the generation efficiency of the point cloud map to reduce the amount of storage space.The experimental results show that the research method of this paper is feasible in the RGB-D SLAM algorithm,and it can meet the real-time requirements with high accuracy.