NDT使用openmp加速

开源代码地址:[email protected]:koide3/ndt_omp.git

该开源代码,主要测试了普通NDT以及使用openmp加速之后的ndt匹配效率及精度问题。我实际测试了大尺度的点云地图匹配(室外200mx50m),得到的结果是GICP匹配效果差很容易飘(如果想要更深层次测试GICP的话,可以去关注一下git上面fast_gicp这个包),NDT以及NDT_OMP匹配效果比较稳定。

开源代码的固定帧测试结果:

--- pcl::GICP ---
single : 300.916[msec]
10times: 1450.77[msec]
fitness: 0.2203

--- pclomp::GICP ---
single : 221.72[msec]
10times: 3749.15[msec]
fitness: 0.220517			

--- pcl::NDT ---
single : 423.835[msec]
10times: 4351.72[msec]
fitness: 0.213937

--- pclomp::NDT (KDTREE, 1 threads) ---
single : 365.468[msec]
10times: 3429.27[msec]
fitness: 0.213937

--- pclomp::NDT (DIRECT7, 1 threads) ---
single : 150.935[msec]
10times: 1405.41[msec]
fitness: 0.214205

--- pclomp::NDT (DIRECT1, 1 threads) ---
single : 55.7785[msec]
10times: 429.982[msec]
fitness: 0.208511

--- pclomp::NDT (KDTREE, 8 threads) ---
single : 121.142[msec]
10times: 957.075[msec]
fitness: 0.213937

--- pclomp::NDT (DIRECT7, 8 threads) ---
single : 71.5185[msec]
10times: 508.991[msec]
fitness: 0.214205

--- pclomp::NDT (DIRECT1, 8 threads) ---
single : 54.9451[msec]
10times: 329.652[msec]
fitness: 0.208511

你可能感兴趣的:(0_1SLAM,c++,算法,自动驾驶,点云配准)