Probablistic point cloud resitration algorithms

Project description

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

• Maximum likelihood when the target or source point cloud is observation data
• Variational Bayesian inference
• Distance minimization of two probabilistic distributions
• Hierarchical Stocastic model

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
target = copy.deepcopy(source)
# transform target point cloud
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])
```

Results

Compare algorithms

CPD SVR GMMTree FilterReg

Noise test

ICP(Open3D) CPD FilterReg

Non rigid registration

CPD SVR Filterreg BCPD

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},
version = {0.1.6},
date = {2019-9-29},
}
``````

Project details

Uploaded `source`