2. Point Cloud Library(PCL)
Large scale, open project for 2D/3D
image and point cloud processing
cross-platform
Modular
Contains numerous algorithm used to
extract key points, compute descriptors
to recognize objects etc.
3. Automated 3D-Model Identification
using PCL
Point cloud data creation
Key-point Extraction
Descriptors
Matching (KNN)
Point cloud
data
Key-point
Extraction
Descriptor Matching
4. Point Cloud Data
A point cloud is a data
structure used to represent a
collection of multi-
dimensional points
3D-Point Cloud
Code Representation
6. Descriptors/Features
Pointwise descriptors
Simple, efficient, but not robust to
noise, often not descriptive
enough
Local/Regional descriptors
Locality: features are local
Distinctiveness: can differentiate a
large database of objects
Global descriptors
Higher invariance, well suited for
retrieval and categorization
More descriptive on objects with
poor geometric structure
Histogram Based Graph Based
7. Matching
KNN algorithms are
used
K-Nearest
Neighbours
Non-parametric
method for
regression and
classification
K : number of close
neighbours
N(0)>N(a), so c belongs to o
8. References
1.Point Cloud Library. http://pointclouds.org/about/
2. Federico Tombari. June 4, 2013. Keypoints and Features.
http://www.pointclouds.org/assets/uploads/cglibs13_features.pdf
3. Features and Image Matching.
https://courses.cs.washington.edu/courses/cse455/09wi/Lects/lect6.pdf