论文部分内容阅读
本文研究了近几年迅速发展的以深度相机作为传感器的视觉SLAM(simultaneous localization and mapping)。相比于传统视觉SLAM方案中采用的K-d树或者K-means算法,本文提出一种基于K-means++算法的闭环检测方案,修正了系统的累计误差,提高了系统稳定性和定位精度。在系统前端,选用了基于ORB(Oriented FAST and Rotated BRIEF)特征的特征点法。后端包括位姿图优化和闭环检测,位姿图优化借助g2o(general graph optimization)通用求解器来实现,闭环检测采用了基于二维图像特征的词袋库模型。实验结果成功构建了清晰的三维环境点云地图并计算出精确的运动轨迹。
This paper studies the simultaneous localization and mapping (SLAM) using the depth camera as sensor in recent years. Compared with the K-d tree or K-means algorithm used in the traditional vision SLAM scheme, a closed-loop detection scheme based on K-means ++ algorithm is proposed in this paper, which can correct the cumulative error of the system and improve the system stability and positioning accuracy. At the front of the system, the feature point method based on the ORB (Oriented FAST and Rotated BRIEF) feature is selected. The back end includes pose optimization and closed-loop detection. The pose optimization is realized by using general solver g2o (general graph optimization). The closed-loop detection uses the bag-of-words model based on two-dimensional image features. The experimental results successfully constructed a clear three-dimensional environment point cloud map and calculate the precise trajectory.