Point Cloud Processing

Working with 3D point clouds: representation, filtering, segmentation, registration, and surface reconstruction for robotic perception.

Learning Objectives

1. Point Cloud Representation

1.1 XYZ / XYZRGB / XYZI

1.2 Common Formats (PCD, PLY, LAS)

2. Point Cloud Acquisition

2.1 From RGB-D Camera

2.2 From LiDAR

2.3 From Stereo Vision

3. Filtering & Downsampling

3.1 Voxel Grid Filter

3.2 Statistical Outlier Removal

3.3 Pass-Through Filter

4. Segmentation

4.1 RANSAC Plane Segmentation

4.2 Euclidean Cluster Extraction

4.3 Region Growing

5. Registration (ICP & Variants)

5.1 Iterative Closest Point (ICP)

5.2 Point-to-Plane ICP

5.3 Feature-Based Registration (FPFH)

6. Surface Reconstruction

7. 3D Object Recognition

8. Libraries: Open3D & PCL

Exercises

References