Skip to content
/ probreg Public
forked from neka-nat/probreg

Python package for point cloud registration using probabilistic model (Coherent Point Drift, GMMReg, SVR, GMMTree, FilterReg, Bayesian CPD)

License

Notifications You must be signed in to change notification settings

OOub/probreg

 
 

Repository files navigation

logo

Build Status Build status PyPI version MIT License Documentation Status Downloads

Probreg is a library that implements point cloud registration algorithms with probablistic model.

The point set registration algorithms using stochastic model are more robust than ICP(Iterative Closest Point). This package implements several algorithms using stochastic models and provides a simple interface with Open3D.

Core features

  • Open3D interface
  • Rigid and non-rigid transformation

Algorithms

Transformations

type CPD SVR, GMMReg GMMTree FilterReg BCPD (experimental)
Rigid Scale + 6D pose 6D pose 6D pose 6D pose
(Point-to-point,
Point-to-plane,
FPFH-based)
-
NonRigid Affine, MCT TPS - Deformable Kinematic
(experimental)
Combined model
(Rigid + Scale + NonRigid-term)

CUDA support

You need to install cupy.

pip install cupy

Installation

You can install probreg using pip.

pip install probreg

Or install probreg from source.

git clone https://github.com/neka-nat/probreg.git --recursive
cd probreg
pip install -e .

Getting Started

This is a sample code that reads a PCD file and calls CPD registration. You can easily execute registrations from Open3D point cloud object and draw the results.

import copy
import numpy as np
import open3d as o3
from probreg import cpd

# load source and target point cloud
source = o3.io.read_point_cloud('bunny.pcd')
target = copy.deepcopy(source)
# transform target point cloud
th = np.deg2rad(30.0)
target.transform(np.array([[np.cos(th), -np.sin(th), 0.0, 0.0],
                           [np.sin(th), np.cos(th), 0.0, 0.0],
                           [0.0, 0.0, 1.0, 0.0],
                           [0.0, 0.0, 0.0, 1.0]]))
source = source.voxel_down_sample(voxel_size=0.005)
target = target.voxel_down_sample(voxel_size=0.005)

# compute cpd registration
tf_param, _, _ = cpd.registration_cpd(source, target)
result = copy.deepcopy(source)
result.points = tf_param.transform(result.points)

# draw result
source.paint_uniform_color([1, 0, 0])
target.paint_uniform_color([0, 1, 0])
result.paint_uniform_color([0, 0, 1])
o3.visualization.draw_geometries([source, target, result])

Resources

Results

Compare algorithms

CPD SVR GMMTree FilterReg

Noise test

ICP(Open3D) CPD FilterReg

Non rigid registration

CPD SVR Filterreg BCPD

Feature based registration

FPFH FilterReg

Time measurement

Execute an example script for measuring time.

OMP_NUM_THREADS=1 python time_measurement.py

# Results [s]
# ICP(Open3D):  0.02030642901081592
# CPD:  3.6435861150093842
# SVR:  0.5795929960149806
# GMMTree:  0.34479290700983256
# FilterReg:  0.039795294986106455

Citing

@software{probreg,
    author = {{Kenta-Tanaka et al.}},
    title = {probreg},
    url = {https://probreg.readthedocs.io/en/latest/},
    version = {0.1.6},
    date = {2019-9-29},
}

About

Python package for point cloud registration using probabilistic model (Coherent Point Drift, GMMReg, SVR, GMMTree, FilterReg, Bayesian CPD)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 73.5%
  • C++ 26.4%
  • Makefile 0.1%