How pointcloudlibrary/pcl works
PCL is a standalone library for 2D/3D image and point-cloud processing. A typical user supplies point-cloud or sensor data, processes it through representation, search and analysis algorithms, then consumes results in applications or visualization. The tree also shows distinct algorithm domains and GPU, stereo, and sensor capabilities. Most algorithm relationships are inferred from product structure rather than inspected call sites, so those links are dashed; the OpenNI device-manager relationship is visible in the sampled source.
Main entry points: point_cloud.h Cloud data model · features Feature extraction · io Data I/O · openni2_device_manager.cpp OpenNI2 devices
Components 21
Data and I/O
- Data I/O
io - OpenNI2 devices
io/src/openni2/openni2_device_manager.cpp - OpenNI2 device
io/include/pcl/io/openni2/openni2_device.h - RealSense devices
io/src/real_sense/real_sense_device_manager.cpp - Cloud data model
common/include/pcl/point_cloud.h
Point Processing
- Common operations
common/include/pcl/common - Cloud filters
filters - Spatial search
search/include/pcl/search/search.h - Feature extraction
features - Keypoint detection
keypoints
Analysis and Modeling
- Registration
registration - Segmentation
segmentation - Recognition
recognition - Surface modeling
surface - Stereo matching
stereo
Visualization and Apps
- Visualization
visualization - User applications
apps
Specialized Processing
- GPU algorithms
gpu - CUDA search
cuda/nn/organized_neighbor_search.hpp - Machine learning
ml
Other components
- Library user
Connections 23
- Library user → Data I/O: supplies data
- Data I/O → Cloud data model: loads clouds
- OpenNI2 devices → OpenNI2 device: creates device
io/src/openni2/openni2_device_manager.cpp - Library user → OpenNI2 devices: selects sensor
- RealSense devices → Library user: returns capture
- Cloud data model → Common operations: processes geometry
- Cloud data model → Cloud filters: filters points
- Cloud data model → Spatial search: queries neighbors
- Spatial search → Feature extraction: supports estimation
- Cloud data model → Feature extraction: extracts descriptors
- Cloud data model → Keypoint detection: detects points
- Feature extraction → Registration: matches features
- Cloud data model → Registration: aligns clouds
- Cloud data model → Segmentation: partitions points
- Feature extraction → Recognition: classifies objects
- Cloud data model → Surface modeling: builds surfaces
- Cloud data model → Stereo matching: matches images
- Cloud data model → Visualization: renders clouds
- Library user → User applications: uses tools
- User applications → Visualization: displays results
- Cloud data model → GPU algorithms: accelerates processing
- Cloud data model → CUDA search: searches neighbors
- Feature extraction → Machine learning: provides features