研究目的
To improve the precision of 3D laser SLAM point cloud by presenting an offline coarse-to-fine precision optimization algorithm.
研究成果
The proposed coarse-to-fine precision optimization algorithm successfully improved the precision of the 3D laser SLAM point cloud in various environments. The RMSE of point clouds was reduced by approximately 40%–50% and up to 63% after optimization in different datasets. The algorithm also demonstrated good performance on the Cartographer public dataset, showing its versatility and efficiency.
研究不足
The algorithm's computational efficiency is bottlenecked by the ICP registration process. The similarity matching problem in the repeated structure area could be improved by integrating image data as auxiliary information.
1：Experimental Design and Method Selection:
The algorithm involves segmenting and registering point clouds at the local level, constructing a pose graph using feature similarity and global registration, and aligning all segments into the final optimized result. A cycle based error edge elimination method is used to guarantee the consistency of the pose graph.
2：Sample Selection and Data Sources:
The experimental datasets contain three different environments: an indoor office, an underground garage, and the road around an office building.
3：List of Experimental Equipment and Materials:
A customer hand-held laser SLAM device consisting of a laser scanner and an IMU, and a computer with Intel Core i7-3770K CPU @ 3.4 GHz and 8.0 GB of RAM.
4：4 GHz and 0 GB of RAM.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: The raw point cloud is segmented by time and processed to remove noise. The normal vector of each point is estimated. Then, in each segment, setting the first frame point cloud as reference, all point clouds are registered sequentially by point-to-plane ICP and merged. A pose graph is constructed using the features of each point cloud segment to calculate the similarity between different segments. The point cloud segments with high similarity are registered based on features matching and ICP. A cycle based error edge elimination method is performed to ensure the consistency of the pose graph. Then, the poses of each point cloud segment are computed by solving the pose graph. Finally, all segments are registered using the estimated pose.
5：Data Analysis Methods:
The precision of the point cloud after optimization is evaluated using the RMSE of point-to-point distances and the mean plane-to-plane distances compared to reference data obtained by terrestrial laser scanning (TLS).
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
