LCR-SMM: Large Convergence Region Semantic Map Matching through Expectation Maximization
LCR-SMM is a large convergence region semantic map matching algorithm, with a transformation sampling strategy to reduce the initial error.
- PCL
- Eigen
- Sophus
- CERES
$ git clone https://github.com/zqxbit/lcr-smm
$ cd lcr-smm
$ mkdir build && cd build
$ cmake ../
$ make$ ./lcr-smm -s ../data/00_s_S.pcd -t ../data/00_s_T.pcdEstimated Transformation:
0.865958 0.500116 -0.000395343 -4.00093
-0.500116 0.865958 0.0006292 -6.9226
0.000657024 -0.000347144 1 0.0161345
0 0 0 1pcl_viewer -bc 255,255,255 init.pcd
pcl_viewer -bc 255,255,255 LCR.pcd
$ ./lcr-smm -s Path/to/Source/Map.pcd -t Path/to/Target/Map.pcdThe input semantic maps need to be transformed to three-dimensional point clouds with semantic labels and saved as pcd files.
Ajust the x_inf, x_sup,..., r_sup in the main function to delimit the search scope, which decides the convergence region of the algorithm.
Change the value of num_x, num_y, and num_r will determine the number of sampled transformation matrices, and help you to balance the accuracy and efficency of initial optimization.
If you find our work useful in your research, please consider citing:
@article{2021_LCR_SMM,
title={LCR-SMM: Large Convergence Region Semantic Map Matching through Expectation Maximization},
author={Zhang, Qingxiang and Wang, Meiling and Yue, Yufeng and Liu, Tong},
journal={IEEE/ASME Transactions on Mechatronics},
pages={1--11},
doi = {10.1109/TMECH.2021.3124994},
publisher={IEEE}
}