ndt_omp

repository·master·Indexed 21 days ago

A high-performance, multi-threaded, and SSE-optimized implementation of the Normal Distributions Transform (NDT) and Generalized Iterative Closest Point (GICP) algorithms derived from the Point Cloud Library (PCL). Designed for ROS1, it offers multiple neighbor voxel search methods including KDTREE, DIRECT7, and DIRECT1 to balance accuracy and speed.

Tokens: 736
Snippets: 1
Records: 3
Agent score: 25%

What's inside ndt_omp

Overview of ndt_omp
master
ndt_omp is an OpenMP-boosted implementation of the Normal Distributions Transform (NDT) and Generalized Iterative Closest Point (GICP) algorithms, derived from the Point Cloud Library (PCL).
Key features include:
- Performance: Modified to be SSE-friendly and multi-threaded, capable of running up to 10 times faster than the original pcl::NDT.
- Compatibility: This specific repository is designed for ROS1. For ROS2 or non-ROS projects, use the following forks:
  dfki-ric/pclomp
  tier4/ndt_omp
https://github.com/koide3/ndt_omp/blob/master/README.md

Choose a neighbor voxel search method in ndt_omp

master

The package implements several methods for neighbor voxel search. Choosing the right method depends on your requirements for accuracy versus speed:

Method	Description	Recommendation
`pclomp::KDTREE`	Produces results identical to the original `pcl::NDT`.	Use for maximum compatibility with PCL.
`pclomp::DIRECT7`	Faster and stable.	Recommended for general use.
`pclomp::DIRECT1`	Extremely fast registration but potentially unstable.	Use only when speed is the absolute priority.

Benchmark results for ndt_omp

master

Performance benchmarks conducted on a Core i7-6700K comparing pcl::NDT against various pclomp::NDT configurations. Results show significant speedups when using multi-threading and the DIRECT search methods.

$ roscd ndt_omp/data
$ rosrun ndt_omp align 251370668.pcd 251371071.pcd
--- pcl::NDT ---
single : 282.222[msec]
10times: 2921.92[msec]
fitness: 0.213937

--- pclomp::NDT (KDTREE, 1 threads) ---
single : 207.697[msec]
10times: 2059.19[msec]
fitness: 0.213937

--- pclomp::NDT (DIRECT7, 1 threads) ---
single : 139.433[msec]
10times: 1356.79[msec]
fitness: 0.214205

--- pclomp::NDT (DIRECT1, 1 threads) ---
single : 34.6418[msec]
10times: 317.03[msec]
fitness: 0.208511

--- pclomp::NDT (KDTREE, 8 threads) ---
single : 54.9903[msec]
10times: 500.51[msec]
fitness: 0.213937

--- pclomp::NDT (DIRECT7, 8 threads) ---
single : 63.1442[msec]
10times: 343.336[msec]
fitness: 0.214205

--- pclomp::NDT (DIRECT1, 8 threads) ---
single : 17.2353[msec]
10times: 100.025[msec]
fitness: 0.208511