基于局部块特征优化的室内高精度激光SLAM方法

Indoor High-Precision Laser SLAM Based on Local Block Feature Optimization

针对目前激光即时定位与地图构建(Simultaneous Localization and Mapping, SLAM)算法在面对几何特征匮乏的室内环境时易产生定位累积误差大、建图效果较差的问题,提出了基于局部块特征优化并融合惯性测量单元(Inertial Measurement Unit, IMU)数据的激光SLAM方法,在激光点云位置信息和强度信息的深度图上划分局部块作为最小单元,并提取几何域特征和强度域特征作为匹配依据,同时融合IMU数据计算帧间位姿变换,改善了几何特征匮乏环境的建图效果并提升了轨迹计算精度。最后,通过该方法与先进的LOAM和LeGO-LOAM方法在KITTI数据集的仿真实验结果比对以及真实环境建图效果分析,证明局部块几何特征有效减少了累积误差,提升了匹配精度,同时局部块强度特征对几何特征匮乏的情况增加了强度域的约束,改善了“跑飞”等不良现象。

In view of the simultaneous localization and mapping(SLAM)algorithm in the face of indoor environment which lacks geometric features is easy to produce positioning cumulative error and poor building figure effect,a laser SLAM method based on local block feature optimization and fusion of inertial measurement unit(IMU)data is proposed.A local block is divided as the minimum unit on the depth map of laser point cloud location information and intensity information,and the feature of geometric domain and intensity domain is extracted as matching basis.At the same time,IMU data is integrated to calculate the pose transformation between frames,which improves the mapping effect and trajectory calculation accuracy in the environment with insufficient geometric features.Finally,by contrasting the simulation results of the method and advanced LOAM and LeGO-LOAM method on KITTI data sets and real environment building figure effect analysis,it is proved that local geometric feature effectively reduces the cumulative error and improves the matching accuracy,and local intensity feature adds constraints on intensity domain,the“lose control”and other undesirable phenomena are improved.