matlab-register-point-clouds

name	matlab-register-point-clouds
description	Register 3-D point clouds using ICP, NDT, LOAM, FGR, phase correlation, and CPD algorithms. Use when registering or aligning 3-D point clouds, choosing a registration algorithm, tuning registration parameters, preprocessing point clouds for registration or combining point clouds after registration.
license	https://www.mathworks.com/content/dam/mathworks/license/pmrl/license.md
metadata	{"author":"MathWorks","version":"2.0"}

Point Cloud Registration

Select pcregistericp with Metric="planeToPlane" as the starting point, evaluate algorithm accuracy using RMSE, and align the point clouds using pcalign.

When to Use

• User asks to register, align or match 3-D point clouds represented as pointCloud objects
• User needs to choose or tune a registration algorithm
• User wants to build a map from multiple lidar scans
• User wants to combine or merge registered point clouds
• User asks about initial transformation estimation for registration
• User asks to preprocess point clouds before registration

When NOT to Use

• User wants to register 2-D point clouds (use matchScans, matchScansGrid, or matchScansLine)
• User wants image registration (use imregtform or imregcorr from Image Processing Toolbox)

Required Toolboxes

• Computer Vision Toolbox
• Lidar Toolbox

Legacy Patterns to Avoid

Do NOT use	Use instead	Why
rigid3d for InitialTransform	rigidtform3d	rigidtform3d uses the premultiply transformation convention, which is more commonly used.

Algorithm Selection Guide

Local Registration

If you have an initial transformation or the point clouds are only slightly misaligned, try these local registration methods in this order. Use the RMSE output as the accuracy metric to minimize.

Algorithm	Function	Notes
Generalized ICP (plane-to-plane)	pcregistericp with Metric="planeToPlane"	Default starting point. Additionally, try "planeToPlaneWithColor" if the point cloud has color.
ICP (point-to-plane)	pcregistericp with Metric="pointToPlane"	Additionally, try "pointToPlaneWithColor" if the point cloud has color.
ICP (point-to-point)	pcregistericp with Metric="pointToPoint"	Simplest metric. Try if plane-based metrics are inaccurate.
LOAM	pcregisterloam	Requires organized point clouds. If the point cloud is not organized, but you have the lidar parameters available, use pcorganize to organize the point cloud. Use detectLOAMFeatures first to tune feature detection. It gives you more flexibility on the accuracy vs speed tradeoff since detectLOAMFeatures has parameters to control the number of features.
NDT	pcregisterndt	Tune the gridStep parameter for the scale of the scene.

Ground Data Registration

If you don't have an initial transformation, but you know that the data corresponds to ground data, try phase correlation.

Algorithm	Function	Notes
Phase Correlation	pcregistercorr	Specifically designed for ground data. Not a local registration method.

Global Registration

If there is no initial transformation and the point clouds are significantly misaligned (overlap less than 50%), first use a global registration approach to align the moving point cloud to the fixed point cloud, then optionally use a local registration method as a refinement step to improve accuracy. Note that global registration approaches are usually slower than local ones.

Algorithm	Function	Notes
FGR	pcregisterfgr	Feature-based approach.
CPD (rigid)	pcregistercpd with Transform="Rigid"	Probabilistic framework. It handles outliers well. Use for small point clouds.
CPD (nonrigid)	pcregistercpd with Transform="Nonrigid"	Only option for non-rigid/deformable registration.
FPFH feature matching	extractFPFHFeatures + pcmatchfeatures + estgeotform3d	Extract FPFH descriptors, match features, and estimate the transformation using the MSAC algorithm (a variant of RANSAC). Tune NumNeighbors in extractFPFHFeatures and gridStep in pcdownsample to ensure enough features and feature matches.

General Guidelines

• The initial transformation is the transformation that aligns the moving point cloud to the fixed point cloud. It is important to set the InitialTransform name-value argument for local registration algorithms when an estimate is available. If registering a sequence of point clouds and no other estimate is available, assume constant speed and set the initial transformation to the transformation estimated between the previous pair of point clouds.
• The RMSE output of pcregisterfgr is not comparable to the RMSE output of the other registration functions because it is the root mean squared error between only inlier points. Use the Point Cloud Registration Analyzer app for comparable RMSE values between all registration approaches.
• When comparing registration algorithms, check if the point clouds are organized (i.e., ndims(ptCloud.Location) == 3). If they are organized, include LOAM in the comparison. If they are not organized, but lidar parameters are available (e.g., from the sensor metadata or the user), use pcorganize to organize the point clouds and include LOAM in the comparison. Exclude LOAM when the point clouds are unorganized and no lidar parameters are available.
• When registering a sequence of organized point clouds with LOAM, use pcmaploam to refine the transformations and improve map accuracy in addition to pairwise pcregisterloam registration.
• Use pcshowpair to visualize the alignment between a pair of registered point clouds.

Preprocessing Steps

• When using select to filter points from an organized point cloud, always set OutputSize="full" to preserve the organized structure.
• To improve the speed and accuracy of the registration, use findPointsInCylinder as a cylindrical filter to remove artifacts from the ego vehicle and noise from distant points. Only use findPointsInCylinder for point clouds from spinning lidar sensors.
• Downsampling can speed up registration and improve accuracy by removing the effects of noisy points. Use pcdownsample to downsample the point clouds before registration. The recommended approach for registration is the gridAverage or the gridNearest methods. gridAverage is faster and gridNearest returns more accurate color, normal, and intensity values if available in the point cloud.
• Ground removal is not recommended since it can introduce misalignment in the XY plane. In some scenarios it helps speed up registration and improve accuracy, but first evaluate registration results without ground removal. Then, use either segmentGroundSMRF for more accurate ground removal or pcfitplane to get a faster approximation of the ground points.

Parameter Tuning

When tuning parameters, try a range of values between the typical values listed below (choose a step so that it does not exceed 10 values to try). Then, narrow it down further by trying values closer to the best result from that round. The best result corresponds to the lowest RMSE.

Preprocessing Functions

Function	Parameter	Typical Range	Tuning Strategy
findPointsInCylinder	radius(2) (max radius)	90%–100% of point cloud extent	Compute extent as max(abs([ptCloud.XLimits, ptCloud.YLimits])). Try without findPointsInCylinder too.
findPointsInCylinder	radius(1) (min radius)	0–10	Must be less than radius(2). Keep small until it fully removes the ego vehicle.
pcdownsample	gridStep	0.01–1.0

Registration Functions

Function	Parameter	Typical Range	Tuning Strategy
pcregistericp	Metric	All metric values	Try each metric.
pcregistericp	InlierRatio	0.5–1.0 (default: 1)
pcregisterndt	gridStep	0.01–1.5
pcregisterndt	OutlierRatio	0.1–1.0 (default: 0.55)
pcregisterloam / LOAMPoints / downsampleLessPlanar	gridStep	0.1–2.0
pcregisterloam / detectLOAMFeatures	NumRegionsPerLaser	4-15 (default: 6)	Try increasing to extract more features. The value must be an integer.
pcregisterloam / detectLOAMFeatures	MaxSharpEdgePoints, MaxPlanarSurfacePoints	Above defaults of 1	Try increasing to extract more features. The value must be an integer.
pcmaploam	voxelSize	0.1–2.0
pcmaploam / findPose	SearchRadius	2–5 (default: 3)
pcregisterfgr	gridSize	0.01–1.5
pcregistercpd	OutlierRatio	0.1–0.9 (default: 0.1)
pcregistercorr	gridSize	50–250
pcregistercorr	gridStep	0.01–1.5
pcmatchfeatures

Combining Point Clouds

• If you found transformations that align a sequence of point clouds, use pcalign to combine them efficiently. pcalign applies a grid filter across the resulting point cloud to eliminate duplicate points, producing a uniformly dense point cloud.
• If you are combining point clouds that were already transformed using pctransform and you need to preserve every point without any downsampling, use pccat. pccat is the fastest approach since it only concatenates points, but results in a larger point cloud with potential duplicates that can affect the speed and accuracy of downstream workflows.
• If you are combining just 2 point clouds at a time that were already transformed using pctransform and you want to downsample only the region of overlap to remove duplicate points, use pcmerge. Note that the density of the resulting point cloud may not be uniform since downsampling is only applied to the overlap area. pcalign is recommended over pcmerge because it is faster, supports more than 2 point clouds, produces uniform density, and performs the alignment of the point clouds using the transformation from registration.

Interactive Comparison

Use the Point Cloud Registration Analyzer app to visually compare algorithms, tune parameters, and analyze results interactively:

pointCloudRegistrationAnalyzer

The app supports ICP, NDT, LOAM, FGR, Phase Correlation, and CPD. It provides preprocessing (ROI, downsample, ground removal), side-by-side comparison, and it exports the results to the workspace. It supports registration of a pair of point clouds and currently, it does not support the planeToPlaneWithColor and pointToPlaneWithColor metrics of ICP.

Patterns

Basic Registration

fixedPtCloud = pcread("scan1.pcd");
movingPtCloud = pcread("scan2.pcd");

[tform,movingReg,rmse] = pcregistericp(movingPtCloud,fixedPtCloud,Metric="planeToPlane");
pcshowpair(movingReg,fixedPtCloud)

Preprocessing: Cylindrical Filter and Downsample

radius = [4 100];
idx = findPointsInCylinder(ptCloud,radius);
ptCloudFiltered = select(ptCloud,idx,OutputSize="full");
gridStep = 0.1;
ptCloudDownsampled = pcdownsample(ptCloudFiltered,"gridAverage",gridStep);

Registration with Preprocessing

fixedPtCloud = pcread("scan1.pcd");
movingPtCloud = pcread("scan2.pcd");

% Preprocess
radius = [4 100];
gridStep = 0.1;
fixedIdx = findPointsInCylinder(fixedPtCloud,radius);
fixedFiltered = select(fixedPtCloud,fixedIdx,OutputSize="full");
fixedDownsampled = pcdownsample(fixedFiltered,"gridAverage",gridStep);
movingIdx = findPointsInCylinder(movingPtCloud,radius);
movingFiltered = select(movingPtCloud,movingIdx,OutputSize="full");
movingDownsampled = pcdownsample(movingFiltered,"gridAverage",gridStep);

% Register
[tform,movingReg,rmse] = pcregistericp(movingDownsampled,fixedDownsampled,Metric="planeToPlane");
pcshowpair(movingReg,fixedDownsampled)

Global-to-Local Refinement with FGR and ICP

gridSize = 0.5;
tformGlobal = pcregisterfgr(movingPtCloud,fixedPtCloud,gridSize);

alignedPtCloudGlobal = pctransform(movingPtCloud,tformGlobal);

[tform,movingReg,rmse] = pcregistericp(alignedPtCloudGlobal,fixedPtCloud, ...
    Metric="planeToPlane");

pcshowpair(movingReg,fixedPtCloud)

NDT Registration

fixedPtCloud = pcread("scan1.pcd");
movingPtCloud = pcread("scan2.pcd");

gridStep = 0.5;
[tform,movingReg,rmse] = pcregisterndt(movingPtCloud,fixedPtCloud,gridStep);
pcshowpair(movingReg,fixedPtCloud)

LOAM with Feature Detection

movingFeatures = detectLOAMFeatures(movingOrganized);
fixedFeatures = detectLOAMFeatures(fixedOrganized);

gridStep = 1;
movingFeatures = downsampleLessPlanar(movingFeatures,gridStep);
fixedFeatures = downsampleLessPlanar(fixedFeatures,gridStep);

[tform,rmse] = pcregisterloam(movingFeatures,fixedFeatures);
alignedMovingOrganized = pctransform(movingOrganized,tform);

pcshowpair(alignedMovingOrganized,fixedOrganized)

Phase Correlation for Ground Data

fixedPtCloud = pcread("scan1.pcd");
movingPtCloud = pcread("scan2.pcd");

% Align ground planes to the X-Y plane
maxDistance = 0.4;
referenceVector = [0 0 1];
groundFixed = pcfitplane(fixedPtCloud,maxDistance,referenceVector);
groundMoving = pcfitplane(movingPtCloud,maxDistance,referenceVector);
tformFixed = normalRotation(groundFixed,referenceVector);
tformMoving = normalRotation(groundMoving,referenceVector);
fixedCorrected = pctransform(fixedPtCloud,tformFixed);
movingCorrected = pctransform(movingPtCloud,tformMoving);

% Register using phase correlation
gridSize = 100;
gridStep = 0.5;
[tform,rmse] = pcregistercorr(movingCorrected,fixedCorrected,gridSize,gridStep);

movingReg = pctransform(movingPtCloud,rigidtform3d(tformFixed.A * tformMoving.A * tform.A));
pcshowpair(movingReg,fixedPtCloud)

FPFH Feature-Based Registration

fixedPtCloud = pcread("scan1.pcd");
movingPtCloud = pcread("scan2.pcd");

% Downsample
gridStep = 0.5;
fixedDownsampled = pcdownsample(fixedPtCloud,"gridAverage",gridStep);
movingDownsampled = pcdownsample(movingPtCloud,"gridAverage",gridStep);

% Extract FPFH features
neighbors = 40;
[fixedFeature,fixedValidInds] = extractFPFHFeatures(fixedDownsampled, ...
    NumNeighbors=neighbors);
[movingFeature,movingValidInds] = extractFPFHFeatures(movingDownsampled, ...
    NumNeighbors=neighbors);

fixedValidPtCloud = select(fixedDownsampled,fixedValidInds);
movingValidPtCloud = select(movingDownsampled,movingValidInds);

% Match features
indexPairs = pcmatchfeatures(movingFeature,fixedFeature,movingValidPtCloud, ...
    fixedValidPtCloud,Method="Exhaustive",MatchThreshold=1,RejectRatio=0.99);

matchedFixedPtCloud = select(fixedValidPtCloud,indexPairs(:,2));
matchedMovingPtCloud = select(movingValidPtCloud,indexPairs(:,1));

% Estimate the transformation that aligns the matched pairs
tform = estgeotform3d(matchedMovingPtCloud.Location, ...
    matchedFixedPtCloud.Location,"rigid",MaxDistance=2,MaxNumTrials=3000);

movingReg = pctransform(movingPtCloud,tform);
pcshowpair(movingReg,fixedPtCloud)

Combining Registered Point Clouds

ptCloudArr = [ptCloud1 ptCloud2 ptCloud3 ptCloud4];
tforms = [tform1 tform2 tform3 tform4];
gridStep = 0.01;
ptCloudAligned = pcalign(ptCloudArr,tforms,gridStep);

Incremental Map Building

ptCloudFiles = dir("scans/*.pcd");
numPtClouds = numel(ptCloudFiles);

tforms = repmat(rigidtform3d(),numPtClouds,1);
ptCloudArr = repmat(pointCloud(zeros(0,3)),numPtClouds,1);

% Preprocessing parameters
radius = [4 100];
downsampleGridStep = 0.1;

% Load and preprocess first frame
ptCloud = pcread(fullfile(ptCloudFiles(1).folder,ptCloudFiles(1).name));
idx = findPointsInCylinder(ptCloud,radius);
ptCloudFiltered = select(ptCloud,idx);
ptCloudArr(1) = pcdownsample(ptCloudFiltered,"gridAverage",downsampleGridStep);

initTform = rigidtform3d();

for i = 2:numPtClouds
    % Load and preprocess
    ptCloud = pcread(fullfile(ptCloudFiles(i).folder,ptCloudFiles(i).name));
    idx = findPointsInCylinder(ptCloud,radius);
    ptCloudFiltered = select(ptCloud,idx);
    ptCloudArr(i) = pcdownsample(ptCloudFiltered,"gridAverage",downsampleGridStep);

    [relTform,~,rmse] = pcregistericp(ptCloudArr(i),ptCloudArr(i-1),Metric="planeToPlane",InitialTransform=initTform);
    tforms(i) = rigidtform3d(tforms(i-1).A * relTform.A);

    initTform = relTform;
end

gridStep = 0.01;
ptCloudMap = pcalign(ptCloudArr,tforms,gridStep);
pcviewer(ptCloudMap)

LOAM Map Building

voxelSize = 0.1;
loamMap = pcmaploam(voxelSize);
relPose = rigidtform3d();

gridStep = 1;
ptCloud = ptCloudArr(1);
points = detectLOAMFeatures(ptCloud);
points = downsampleLessPlanar(points,gridStep);
addPoints(loamMap,points,rigidtform3d())

tforms = repmat(rigidtform3d(),numPtClouds,1);

for viewId = 2:numPtClouds
   prevPoints = points;
   ptCloud = ptCloudArr(viewId);
   points = detectLOAMFeatures(ptCloud);
   points = downsampleLessPlanar(points,gridStep);

   relPose = pcregisterloam(points,prevPoints,InitialTransform=relPose);

   [absPose,~,rmseRefinement] = findPose(loamMap,points,relPose);
   addPoints(loamMap,points,absPose);
   tforms(viewId) = absPose;
end

ptCloudMap = pcalign(ptCloudArr,tforms,voxelSize);
pcshow(ptCloudMap)

---

Copyright 2026 The MathWorks, Inc.

---