Structure From Multi-Views

DEMO at first

DEMO

Features

this project is modified from Chapter4_StructureFromMotion
non gpu version is better for beginner of learning CV . i.e. compile the project would be easy .
structure from multiple views
use AKAZE FeatureDetector and DescriptorExtractor instead of ORB
non opencv3_contrib verison

AKAZE VS ORB

We have concluded that, although ORB is faster to compute, AKAZE shows a better compromise between speed and performance than ORB for images with low resolution. Keywords: ORB, AKAZE, UAV, visual odometry, feature matching.

For more details – > paper

Matcher : FLANN-BASED MATCHER VS BFMATCHER

BFMatcher is going to try all the possibilities (which is the meaning of “Brute Force” and hence it will find the best matches.

FLANN, meaning “Fast Library for Approximate Nearest Neighbors”, will be much faster but will find an approximate nearest neighbors. It will find a good matching, but not necessarily the best possible one. You can play with FLANN’s parameters in order to increase the precision (i.e. the “quality” of the matchings), but it will be at the cost of slowing the algorithm.

In other words: FLANN is much faster than BFMatcher but it only finds an approximate nearest neighbor, which is a good matching but not necessarily the best. You can play with the parameters of FLANN in order to increase its speed or its precision.

For more details – > Q&A