Table of contents
Contents
Nineteen chapters across five parts, from the physics of a laser pulse to neural rendering. 4 are free to read in full, and the rest open their first section as a preview.
Foundations
Introduction to 3D Point Clouds
What a point cloud is, where it comes from, and the pipeline this book follows.
LiDAR Principles
The physics of laser ranging, from the pulse outward to the georeferenced point.
Acquisition Systems
Airborne, terrestrial, mobile, UAV, handheld: how to choose, and what each one quietly fails at.
Data Formats and Spatial Indexing
LAS, LAZ, E57, COPC, and the kd-tree / octree machinery underneath them.
Preprocessing
Sampling and Resampling
Voxel-grid, farthest-point, Poisson-disk: when each one is the right answer.
Filtering and Denoising
Statistical, radius, and learned outlier removal, evaluated on real scans.
Normal Estimation and Curvature
PCA-based normal estimation, orientation, and what curvature actually measures.
Registration
Pairwise and global alignment: ICP and its modern variants, plus learned matchers.
Core Processing
Ground Extraction
Cloth simulation, progressive TIN, and how each fails under dense vegetation.
Local Features and Keypoints
Detecting repeatable points across scans before describing them.
Segmentation and Clustering
RANSAC, region growing, DBSCAN, and the geometry of separating one object from the next.
Classification
From hand-tuned rules to random forests to PointNet, with honest accuracy numbers.
Feature Descriptors
SHOT, FPFH, and learned descriptors as the front door to registration and classification.
Advanced Processing
Change Detection
Cloud-to-cloud, cloud-to-mesh, and the M3C2 algorithm for terrain monitoring.
Surface Reconstruction
Poisson, alpha shapes, ball-pivoting, and modern implicit reconstructions.
Object Reconstruction
From segmented points to BIM-ready geometry, with a focus on buildings.
Frontiers
Applications
Where the algorithms go to work: mapping, infrastructure, heritage, autonomous systems.
Deep Learning on Point Clouds
PointNet to PointNeXt and the architectures that actually generalise across datasets.
Neural Rendering and Gaussian Splatting
NeRF, 3DGS, and the new boundary between point clouds and view synthesis.
Want the complete book?
You get an email when it ships.