回环检测作为同步建图与定位(Simulation Localization and Mapping,SLAM)算法中的基本组成部分,能有效关联相同场景之间的特征信息,提供全局一致性的位姿估计.基于词袋(Bag of Words,BoW)模型的回环检测算法在视觉SLAM领域有着显著成效,但对于激光雷达SLAM算法,主流的方法无法实时有效地识别回环场景,且通常无法校正完整的六自由度(6 Degree of Freedom,6-DOF)环路姿态.针对以上问题,文章提出了一种基于线性关键点特征表示的词袋模型,用于激光雷达SLAM中的实时回环检测.该词袋模型计算性能高效,可满足自动驾驶实时性要求.同时,算法具有稳定的姿态校正能力,可用于精确的点对点匹配.在公开数据集上,将文章提出的方法嵌入激光SLAM算法中进行闭环性能评估.结果表明,基于词袋模型的回环检测算法在激光SLAM领域优于现有的主流方法.
Research on Bag of Words loop detection algorithm based on lidar
As a basic component of the Simulation Localization and Mapping(SLAM)algorithm,loop detection can effectively correlate feature information between the same scenes and provide globally consistent pose estimation.The loop detection algorithm based on the Bag of Words(BoW)model has achieved remarkable results in the field of visual SLAM.However,for the lidar SLAM algorithm,the mainstream method cannot effectively identify the loop scene in real time,and usually cannot correct the complete six degrees of freedom loop stance.In response to the above problems,this paper proposes a bag-of-word model based on linear key point feature representation for real-time loop detection in lidar SLAM.The bag-of-words model has efficient computing performance and can meet the real-time requirements of autonomous driving.At the same time,the algorithm has stable attitude correction capabilities and can be used for accurate point-to-point matching.On public data sets,the method proposed in this article is embedded into the laser SLAM algorithm for closed-loop performance evaluation.The results show that the loop detection algorithm based on the bag-of-words model is superior to the existing mainstream methods in the field of laser SLAM.
Bag of Wordsloop detectionlinear key point featuresreal-time
NETL
NSTL
万方数据
