FPGA implementation of feature detection and matching using ORB

Oriented FAST and Rotated BRIEF (ORB) is becoming the method of choice in real-time feature matching applications by offering performances comparable to other state-of-the-art descriptors at a fraction of the runtime and memory requirements. These characteristics have prompted multiple researchers t...

Full description

Saved in:
Bibliographic Details
Published inMicroprocessors and microsystems Vol. 94; p. 104666
Main Authors Belmessaoud, Nadir Mohamed, Bentoutou, Youcef, Chikr El-Mezouar, Miloud
Format Journal Article
LanguageEnglish
Published Elsevier B.V 01.10.2022
Subjects
Online AccessGet full text
ISSN0141-9331
1872-9436
DOI10.1016/j.micpro.2022.104666

Cover

More Information
Summary:Oriented FAST and Rotated BRIEF (ORB) is becoming the method of choice in real-time feature matching applications by offering performances comparable to other state-of-the-art descriptors at a fraction of the runtime and memory requirements. These characteristics have prompted multiple researchers to propose FPGA-based accelerators that allow for a significant increase in throughput compared to general-purpose processors. This increase is, however, achieved at the expense of the descriptor’s robustness as several approximations are necessary to map the algorithm efficiently into hardware which negatively affects its invariance. In this paper, we introduce an FPGA architecture capable of extracting ORB descriptors at high throughput with minimal impact to their invariance. Rotation of sample point coordinates is bypassed entirely, and methods for the accurate computation of the moments and descriptor generation in a single clock cycle are described. Moreover, a feature matching algorithm based on the cross-check criterion is proposed and implemented in hardware. The architecture achieved a throughput of 94 MPix/s using a 100 MHz clock while producing an almost identical performance compared to the reference implementation. •An FPGA-based accelerator of the ORB descriptor is proposed.•Methods for the accurate computation of the patch moments are described.•Rotation of sample point coordinates is bypassed.•The descriptor is computed in a single clock cycle.•Feature matching using the cross-check criterion has been implemented in hardware.
ISSN:0141-9331
1872-9436
DOI:10.1016/j.micpro.2022.104666