基于地面约束和主成分分析特征提取的室内激光SLAM系统

An Indoor Lidar SLAM Based on Ground Constraint and Principal Component Analysis Based Feature Extraction

针对室内同步定位和建图(SLAM)存在特征点稳定性不足及垂直方向误差累积的问题,提出了一种基于地面约束和主成分分析特征提取的室内激光SLAM系统。首先,通过主成分分析方法提取出代表性强、稳定性强的特征点,从而提高特征匹配和位姿优化的准确性;然后,在特征提取模块和建图模块分别检测地面,并将地面约束加入到位姿的计算中。实验结果表明:在各种室内环境中,相较于其他激光SLAM方法,在保证实时效率的同时,本文算法可有效地提高定位精度并减小垂直方向的误差。

Aiming at the problem that the feature points extracted are not robust enough and the accumulated vertical drift errors over long-term operation in current LiDAR SLAM,we propose an indoor LiDAR SLAM algorithm based on ground constraint and principal component analysis based feature extraction.First,principal component analysis is applied to extract more discriminative and robust features,which will improve the accuracy of feature association and pose optimization.Second,the ground is detected in the feature extraction module and the mapping module respectively and the ground constraint is added to estimate the pose.Experiments show that compared with other state-of-the-art method,the proposed algorithm can achieve better accuracy and reduce the vertical error without affecting the real-time performance in various indoor environments.