摘要
通过分析现有2D激光SLAM闭环检测算法由于存在匹配信息量不足的弊端,而容易引起在几何局部结构相似或纹理信息单一的场景中误匹配的问题,提出一种基于multiscansto-map扫描匹配改进的闭环检测算法。通过复合连续数帧激光点云,在降低数据的冗余性的同时提高单次匹配的数据信息量,来提高匹配算法在几何环境相似条件下的区分度,并提出几何一致性检验的闭环验证策略防止出现误匹配,同时通过提出自适应里程信息的submap剔除策略在一定程度上提高闭环检测的匹配效率。通过分析仿真实验结果,验证了改进算法可以有效解决几何结构相似的场景中的误匹配问题,提高了闭环检测算法的鲁棒性。
This specification is set for the theses to be published in computer applications and software,including fonts,margins,page size and print area.By analyzing the disadvantage of insufficient matching information,which is easy to cause problems of mismatch in scenes with similar geometric local structures or simple texture information,in the existing loop closure detection algorithm of 2D laser SLAM,this paper proposed an improved loop closure detection algorithm based on multiscans-to-map scan matching method.By compounding a continuous number of laser point clouds,to reduce data redundancy and increase the amount of data information that is matched in a single match,which is in order to improve the differentiation of scan matching algorithm in scenes with similar geometric local structures.In addition,this paper proposes a loop closure verification strategy which is called geometric consistency testing in order to prevent mismatch,and a submap culling strategy for adaptive odometer information in order to improve the scan matching efficiency of loop closure detection.Finally,by simulation experiment,this paper proves that the improved algorithm can effectively solve the mismatch problem and improves the robustness of the detection algorithm.
作者
黄永新
HUANG Yong-xin(School of Logistics Engineering,Wuhan University of Technology,Wuhan 430063,China)
出处
《自动化与仪表》
2020年第6期42-47,共6页
Automation & Instrumentation
