Benchmarking of Feature Detectors and Matchers using OpenCV-Python Wrapper

This paper proposes a comparative analysis of AKAZE, BRISK, KAZE, ORB, and SIFT features detecting algorithms in combination with BF and FLANN feature matching algorithms. The comparative evaluation was implemented using the OpenCV-Python wrapper. In the first phase, for each pair of algorithms, we...

Full description

Saved in:
Bibliographic Details
Published in2021 International Conference on Information Technology and Nanotechnology (ITNT) pp. 1 - 6
Main Authors Golovnin, Oleg, Rybnikov, Dmitry
Format Conference Proceeding
LanguageEnglish
Published IEEE 20.09.2021
Subjects
computer vision
descriptor
Detectors
Feature detection
Feature extraction
Image resolution
local detector
Measurement uncertainty
OpenCV
Structure from motion
Task analysis
Online AccessGet full text
DOI10.1109/ITNT52450.2021.9649278

Cover

More Information
Summary:This paper proposes a comparative analysis of AKAZE, BRISK, KAZE, ORB, and SIFT features detecting algorithms in combination with BF and FLANN feature matching algorithms. The comparative evaluation was implemented using the OpenCV-Python wrapper. In the first phase, for each pair of algorithms, we estimated the time of detection, the number of detected features, the time of matching, and the number of matched features. In the second phase, we estimated the time of feature detection depending on image resolution and the number of features. The results showed the best summary efficiency is achieved when using a pair of ORBFLANN algorithms. The performance of detectors and matchers is determined only by hardware features, image sets, and configuration while using the OpenCV-Python wrapper does not affect performance.
DOI:10.1109/ITNT52450.2021.9649278