PCD_Registration

This is a ros package that register 2 pointclouds with using either Generalized Iterative Closest Point, which is I recommend, or Normal Distrubiton Transform algorithms.

Requirements

• ROS2 Foxy
• PCL Library

Build

cd ${workspace}/src
git clone https://github.com/yvzksgl/PCD_Registration.git
cd ${workspace}
colcon build
source install/setup.bash

Launch

You may want to change algorithm that will be applied. To do that you need to alter launch file in '${workspace}/src/PCD_Registration/launch'.

parameters=[{"input_pcd1_path":"data/capture0001.pcd"},
                        {"input_pcd2_path":"data/capture0002.pcd"},
                        {"out_pcd_path":"data/transformed.pcd"},
                        {"algorithm": "gicp"}]

Only "gicp" and "ndt" allowed for the algorithm parameter. To launch program as pre-configured:

ros2 launch pcd_reg pcd_reg.launch.py

Results

Both of the results, green illustrates source pointcloud, red illustrates target pointclud and blue illustrates output pointcloud. GICP works pretty good but NDT failed at my implementation. It can be enhanced by parameter tuning but it makes NDT very slow in my case. solution may be find in the future.

• GICP
• NDT without Tuning
• NDT with Tuning