Point Cloud Toolbox

MAJOR UPDATE
MATLAB script for reading point cloud data on the left and point cloud data visualized on the right.

Read, Write, and Visualize Point Clouds

Import point clouds from file formats such as LAS/LAZ, PCD, and PLY, or stream point cloud data directly from lidar sensors, including Velodyne® and Ouster®. Visualize and explore point cloud data across datasets of any size using the Point Cloud Analyzer app.

Get Started with Point Cloud Analyzer

Documentation | Examples

A displacement field is applied to a teapot point cloud, separating the geometry into distinct components while preserving the shape of each part.

Process and Analyze Point Clouds

Process and analyze ground-based and aerial point clouds using algorithms for unorganized-to-organized conversion, ground segmentation, clustering, shape fitting, geometric transformation, and feature extraction.

Estimate Transformation Between Two Point Clouds Using Features

Documentation | Examples

Point Cloud Registration Analyzer app shows the comparison of two registration results and preprocessing steps.

Point Cloud Registration and SLAM

Register point clouds using algorithms such as ICP, NDT, and FPFH-based feature matching. Evaluate and compare registration algorithms using the Point Cloud Registration Analyzer app. Implement 3D SLAM workflows for mapping and localization for ground-based and aerial point clouds.

Implement Point Cloud SLAM in MATLAB

Documentation | Examples

Lidar Camera Calibrator app showing checkerboard detection and calibration error plots.

Lidar Camera Calibration

Estimate rigid transformation between a 3D lidar sensor and a camera. Use the estimated rigid transformation matrix between camera and lidar to fuse the data from these two sensors.

Documentation | Examples

Point cloud with overlaid semantic labels for buildings, vegetation, ground, and other classes.

Deep Learning for Point Clouds

Apply deep learning algorithms for semantic segmentation and object detection on point clouds. Train, test, and evaluate performance of these networks. Deploy them on target hardware by generating C/C++ or CUDA code.

Documentation | Examples

Multi-Sensor Labeler app showing synchronized car labels in video and a lidar point cloud.

Point Cloud and Video Labeling

Label point clouds and videos for object detection, semantic segmentation, and lane detection. Use the Multi-Sensor Labeler app with built-in or custom automation algorithms to accelerate the labeling process.

Documentation | Examples

Product Resources:

Documentation Examples Requirements Release notes
研究者がディープラーニングを使ってチーターから着想を得たロボットを設計

“Imagine being in a car and putting your arm out the window. You feel that force. We determined that the cheetah uses this aerodynamic drag to stabilize its body during high-speed turns.”

Amir Patel, ARU director
View more customer stories

Point Cloud Toolbox FAQs

Point Cloud Toolbox provides algorithms, functions, and apps for designing, analyzing, and testing lidar processing systems, including object detection and tracking, semantic segmentation, shape fitting, lidar registration, and obstacle detection.

Point Cloud Toolbox supports point cloud data from lidar sensors, RGB-D cameras, stereo cameras, and mmWave radar. It can also process data generated through visual SLAM, photogrammetry, and structure-from-motion workflows.

Yes. You can import point cloud data from standard file formats such as PCD, PLY, LAS, LAZ, and PCAP or stream data directly from popular lidar sensors, including Ouster, Velodyne, SICK, and Hokuyo lidars, for real-time processing and analysis.

The Point Cloud Analyzer app enables interactive visualization, analysis, and measurement of point clouds, with built-in preprocessing algorithms for ground removal, denoising, median filtering, cropping, and downsampling, and support for custom preprocessing algorithms.

Yes, the Lidar Camera Calibrator app estimates rigid transformation between a 3D lidar sensor and a camera, enabling data fusion between the two sensors.

Yes. Point Cloud Toolbox provides algorithms for point cloud registration, including ICP, NDT, and FPFH-based feature matching. It also supports 3D SLAM workflows for map building and localization using point cloud data.

The Multi-Sensor Labeler app enables you to label point clouds and videos for training and evaluating perception systems. You can create ground truth labels for object detection, semantic segmentation, and lane detection tasks using data from multiple sensors. The app includes built-in automation algorithms and supports custom automation workflows to accelerate the labeling process.

Yes, you can train detection, semantic segmentation, and classification models using deep learning algorithms, such as PointPillars, SqueezeSegV2, PointNet++, and PointSeg, on lidar point clouds.

Most toolbox algorithms support C/C++ code generation for integrating with existing code, desktop prototyping, and deployment, with some algorithms also supporting CUDA code generation.