pcfitcuboid

R2026b

Fit cuboid to 3-D point cloud

Syntax

model = pcfitcuboid(ptCloud)

model = pcfitcuboid(ptCloud,indices)

Description

model = pcfitcuboid(ptCloud) fits a cuboid over the input point cloud data. The function stores the properties of the cuboid in the cuboidModel object, model.

model = pcfitcuboid(ptCloud,indices) fits a cuboid over a selected set of points, indices, in the input point cloud.

Examples

Fit Cuboid over Point Cloud Data

Load the point cloud data into the workspace.

ptCloud = load("drivingLidarPoints.mat").ptCloud;

Select a region of interest for cuboid fitting.

roi = [-10 10 -6 8 -2 2];
idx = findPointsInROI(ptCloud,roi);
ptCloudROI = select(ptCloud,idx);

Visualize the cropped point cloud.

figure
pcshow(ptCloudROI)
title("Cropped Point Cloud")

Segment the ground plane.

maxDistance = 0.5;
[~,idxGround,idxNoGround] = pcfitplane(ptCloudROI,maxDistance);
groundPtCloud = select(ptCloudROI,idxGround);
ptCloudNoGround = select(ptCloudROI,idxNoGround);

Visualize the segmented ground plane.

figure
pcshowpair(groundPtCloud,ptCloudNoGround)
legend("Ground Point Cloud","Non-Ground Point Cloud")
title("Segmented Ground Plane")

Segment the non-ground point cloud into clusters.

minDistance = 1;
[labels,numClusters] = pcsegdist(ptCloudNoGround,minDistance);

Visualize the point cloud and each cluster in a different color.

figure
pcshow(ptCloudROI)
hold on
pcshow(ptCloudNoGround.Location,labels)

Fit a cuboid to each cluster, and visualize each cluster with its respective fitted cuboid.

for i = 1:numClusters
    idx = find(labels == i);
    model = pcfitcuboid(ptCloudNoGround,idx);
    plot(model)
end
title("Clustered Point Cloud with Fitted Cuboids")

Input Arguments

ptCloud — Point cloud

pointCloud object

Point cloud, specified as a pointCloud object.

indices — Indices of selected points

vector of positive integers

Indices of selected points, specified as a vector of positive integers.

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64

Output Arguments

model — Cuboid model

cuboidModel object

Cuboid model, returned as a cuboidModel object. The function returns an empty cuboid model when the input point cloud contains fewer than four points because at least four points are required to fit a cuboid using a convex-hull based approach.

Algorithms

This algorithm uses the convex hull of the points in the input point cloud to compute an oriented bounding cuboid. The algorithm performs these steps:

1. Compute the convex hull of the 3-D points.

2. For each face of the convex hull, compute a candidate cuboid with one face coplanar with the convex hull face.

3. Select the candidate cuboid with the smallest volume.

When the convex hull contains only a small number of faces, the orientation of the true minimum-volume cuboid might not align with any convex hull face. Because the algorithm considers only orientations derived from convex hull faces, it can return a suboptimal bounding cuboid. For example, four non-coplanar points form a tetrahedral convex hull with only four faces, and the optimal cuboid orientation might not align with any of them.

Version History

Introduced in R2020b

R2026b: Use pcfitcuboidalign for previous results

You can now use the pcfitcuboid function to estimate tighter-fitting cuboids that support yaw, pitch, and roll. Use the new pcfitcuboidalign function to fit cuboids aligned to a plane by specifying a normal vector.

In previous releases, pcfitcuboid returned a cuboidModel with orientation [0 0 theta]. To obtain the same cuboid fit in this release, use the pcfitcuboidalign function.

R2026b: C/C++ and GPU Code Generation support removed for this function

C/C++ Code and GPU code generation support removed for this function. Use the pcfitcuboidalign function for C/C++ and GPU Code Generation support.